JPH08317385A - Image coding apparatus and decoding apparatus - Google Patents

Abstract

(57) [Summary] [Object] The present invention uses the fact that the intermediate level value of the region shape / transparency information, which is an input signal, is small and the distribution of pixel values is biased to two levels, Aims to provide an efficient image coding apparatus and decoding apparatus by performing vector quantization and performing binary layered three-level pattern coding on binary information. [Structure] Blocked signal 10 obtained by dividing input signal 1001 which is area shape / transparency information by blocking means 1002.
03 is vector-quantized by vector quantization means 1004,
Index signal 1005 and vector quantization pattern signal 10
Outputs 06 and. The vector quantization pattern signal 1006 is 3
It is converted into a ternary pattern signal 1008 in the binarizing means 1007, further hierarchized in the hierarchizing means 1009, and output as a hierarchized pattern signal 1010. The hierarchical pattern signal 1010 and the index signal 1005 are code string generating means.
It is encoded in 1011 and output as an encoded signal 1012.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding device and an image decoding device for efficiently coding and decoding image region shape information and transparency information when coding an image for storage and transmission. The present invention relates to a chemical conversion device.

[0002]

2. Description of the Related Art In recent years, motion compensation coding based on region division has been studied for ultra low bit rate moving image coding. However, in encoding with region division, it is necessary to perform extra encoding using region shape information as additional information in addition to pixel value information. However, as conventional techniques, only the following two methods have been considered.

(1) The region shape is also orthogonally transformed (discrete cosine transform or the like) like the luminance signal and the color difference signal.

(2) Binary coding (MH, MR, MMR, chain coding, etc.) by regarding the area shape as a binary image.

FIG. 17 shows a block diagram of a conventional image coding apparatus based on orthogonal transform. In the figure, 1 is an input signal which is a screen to be encoded, 2 is a blocking means, 3 is a blocking signal, 4 is an orthogonal transforming means, 5 is a transform coefficient signal, 6 is a variable length coding means, 7 is a code Signal.

The operation of the image coding apparatus configured as described above will be described below. The blocking means 2 divides the input signal 1 which is the screen to be coded into blocks of a certain processing unit and outputs a blocked signal 3. Vertical 1
Often divided into a rectangular block of 6 pixels by 16 pixels and a rectangular block of 8 pixels in the vertical direction and 8 pixels in the horizontal direction. The orthogonal transform means 4 performs orthogonal transform on the blocked signal and outputs a transform coefficient signal 5. As the orthogonal transform, discrete cosine transform, Haar transform, Hadamard transform, Karhunen-Loeve transform, etc. have been studied, but the discrete cosine transform is often selected in consideration of the coding efficiency and the existence of a fast algorithm. . In the variable length coding means 6, the conversion coefficient 5
Is subjected to variable length coding and output as a coded signal 7.

On the other hand, in recent years, hierarchical encoding has been considered in which an image is separated into layers such as a background and a foreground, each layer is encoded separately, and an encoded image is obtained as a composite image thereof. Figure 17
Figure shows the image hierarchy. As shown in the figure, the hierarchical coding referred to here is a method of separately coding an image, a foreground fish and a background, and combining them at the time of decoding. In this hierarchical encoding, it is also necessary to encode transparency information that is a reference for how much the foreground (fish in the figure) is to be combined with the background. This transparency information (also called an α plane) is additional information including area shape information of the foreground to be combined, and a technique for encoding such area shape information with high efficiency is necessary.

Most of the transparency information is 0% or 100% transparency, and there are few intermediate level values such as 23% or 68%. The intermediate level value appears in a translucent object such as a window glass or a boundary portion of the object. FIG. 18 shows an explanatory diagram regarding the intermediate level value of the transparency information. Also in the figure, most are white and black, and the intermediate level values can be seen only at the boundary.

[0009]

However, in order to efficiently encode the transparency information, there are the following problems.

Most of the transparency information is 0% or 100% transparency, and the intermediate level value is small. For this reason, even if a normal method for compressing a grayscale image (orthogonal transformation such as discrete cosine transformation) is performed, efficient coding cannot be performed.

When the binary image coding is used by performing the binarization, the intermediate level value is omitted, and when the images are combined, a large amount of distortion occurs at the boundary portion of the regions.

In view of such a point, the present invention utilizes the fact that the pixel values of the image representing the area shape and the transparency information are biased in binary distribution, and thus utilizes the hierarchical ternary pattern coding and vector quantization. It is an object of the present invention to provide an image coding apparatus and a decoding apparatus that efficiently code region shape information and transparency information by using.

[0013]

According to a first aspect of the present invention, there is provided an image coding apparatus which blocks an input signal, which is an image to be coded, which is a signal representing image transparency information, area shape information, and the like. Blocking means for dividing and outputting as a blocked signal, vector quantization of the blocked signal for each block, vector quantization pattern signal representing the state of vector quantization, and index representing the index of the selected vector A vector quantization means for outputting a signal to the ternaryization means and a code string generation means, respectively, the vector quantization pattern signal, a block to which an index is assigned, a block in which all pixels are the minimum pixel value, all pixels Is divided into three states of the block having the maximum pixel value, and the ternarization means for ternarizing and outputting as a ternary pattern signal; Hierarchical means for collectively hierarchizing a plurality of blocks in which all pixels have a minimum pixel value and a plurality of blocks in which all pixels have a maximum pixel value, and outputting the value pattern signal as a hierarchical pattern signal; The coded pattern signal and the index signal are coded based on a predetermined condition to generate a code string and output the coded signal as a coded signal.

An image coding apparatus according to a second aspect of the present invention is a signal representing the image transparency information, area shape information, and the like. As block output means to the ternarization means and the vector quantization means, and the block signal is vector-quantized in block units, and an index signal representing the index of the selected vector is output to the code string generation means. Vector quantizing means, and the block signal, a block in which all pixels have a minimum pixel value, a block in which all pixels have a maximum pixel value, other blocks
Dividing into three states, ternarizing means for ternarizing and outputting as a ternary pattern signal, and the ternary pattern signal, a plurality of blocks in which all pixels are minimum pixel values and all pixels are maximum pixel values A plurality of blocks are grouped together and hierarchized,
Hierarchizing means for outputting as a hierarchized pattern signal; code string generating means for encoding the hierarchized pattern signal and the index signal based on a predetermined condition to generate a code string and outputting as a coded signal. Composed of.

An image decoding apparatus of a third invention uses a signal encoded by the image encoding apparatus of the first invention and the second invention as an input signal, and outputs a hierarchical pattern decoding signal and an index decoding signal. , A code string decoding means for outputting to the ternary pattern generation means and the vector dequantization means, and a hierarchical structure is reproduced from the hierarchical pattern decoded signal, and expanded into a ternary pattern. A ternary pattern generating means for outputting a ternary pattern decoded signal, a vector for searching and assigning a vector from the index decoded signal, generating a decoding screen for each processing block, and outputting a block decoded signal Quantizing means, using the three-valued pattern decoded signal and the block decoded signal as input signals, determine the position of the vector-quantized block from the three-valued pattern decoded signal. Then, the block decoding signal is assigned to that position, and in the case of blocks other than the vector quantization block, the minimum pixel value or the maximum pixel value is set to all the pixels of the block according to the value of the ternary pattern decoding signal. And a decoded signal generating means for substituting and outputting as a decoded signal.

An image coding apparatus according to a fourth aspect of the present invention is a signal representing the image transparency information, area shape information, etc., which is an input image which is an image to be coded and is divided into blocks to obtain a block signal. Blocking means for outputting as
A binary / multivalue determination means for determining whether the blocking signal is a binary signal or a multilevel signal and outputting a selection control signal to the first selection means and the second selection means; and the block The selection control signal based on the selection control signal.
When the value is binary, the first selecting means outputs the blocked signal to the second selecting means, and when the selection control is multivalued, the vector quantizing means, and the first selecting means outputs the blocked signal. The blocked signal is vector-quantized in block units, and a vector quantization pattern signal representing the state of vector quantization and an index signal representing the index of the selected vector are respectively generated as second selecting means and code string generation. Vector quantizing means to output to the means, based on the selection control signal, if the selection control is binary output the blocking signal as a selection blocking signal, when the selection control is multi-valued index signal Second selection means for outputting as a selected blocking signal, the selected blocking signal, a block in which all pixels have a minimum pixel value, all pixels have a maximum pixel value Lock, divided into three states of other blocks, ternarizing means for ternarizing and outputting as a ternary pattern signal, and the ternary pattern signal, a plurality of blocks in which all pixels are the minimum pixel value Hierarchizing means for collectively hierarchizing a plurality of blocks in which all pixels have the maximum pixel value, and outputting as a hierarchized pattern signal;
The layered pattern signal and the index signal,
It is composed of a code string generation means for performing coding based on a predetermined condition to generate a code string and output it as a coded signal.

An image coding apparatus according to a fifth aspect of the present invention is a signal representing the image transparency information, area shape information, and the like, which is an image to be coded. Blocking means for outputting as
An image determination unit that classifies the blocked signals based on a predetermined condition and outputs a codebook selection control signal to a codebook selection unit, and selects from at least two codebooks based on the codebook selection control signal, Codebook selecting means for outputting a selected codebook, vector quantization of the blocked signal with reference to the selected codebook on a block-by-block basis, and a vector quantization pattern signal representing a state of vector quantization are selected. And an index signal representing the index of the vector, vector quantizing means for respectively outputting to the ternarization means and the code string generating means, the vector quantization pattern signal, the block to which the index is assigned, all pixels are the minimum pixel Value, block with all pixels having the maximum pixel value And a ternarizing means for ternarizing and outputting as a ternary pattern signal, and a ternary pattern signal, a plurality of blocks in which all pixels have a minimum pixel value and a plurality of blocks in which all pixels have a maximum pixel value. Hierarchization means for collectively classifying blocks and outputting as a hierarchical pattern signal, and the hierarchical pattern signal and the index signal are encoded based on a predetermined condition to generate and encode a code string. And a code string generating means for outputting as a signal.

An image coding apparatus according to a sixth aspect of the present invention is a signal representing image transparency information, area shape information, etc., which is an input image which is an image to be coded and is divided into blocks to obtain a block signal. As block output means to the ternarization means and the vector quantization means, and the block signal is vector-quantized in block units, and an index signal representing the index of the selected vector is output to the code string generation means. Vector quantizing means, and the block signal, a block in which all pixels have a minimum pixel value, a block in which all pixels have a maximum pixel value, other blocks
Dividing into three states, ternarizing means for ternarizing and outputting as a ternary pattern signal, and the ternary pattern signal, a plurality of blocks in which all pixels are minimum pixel values and all pixels are maximum pixel values A plurality of blocks are grouped together and hierarchized,
Layering means for outputting a layer 1 pattern signal and a layer 2 pattern signal for each layer, and layer 1 pattern variable length encoding means for variable length encoding the layer 1 pattern signal and outputting it as a layer 1 pattern variable length encoded signal. And the hierarchy
Layer 2 pattern variable length coding means for variable length coding a 2 pattern signal and outputting as a layer 2 pattern variable length coded signal, the layered 1 pattern variable length coded signal, and the layered 2 pattern variable length code The coded signal and the index signal are coded based on a predetermined condition to generate a code string and output the coded signal as a coded signal.

An image coding apparatus according to a seventh aspect of the present invention is a signal representing the image transparency information, area shape information, and the like, which is an image to be coded and is divided into blocks for each processing unit. As a block quantizing means for outputting to the ternary quantizing means and the vector quantizing means, a vector quantizing means for vector quantizing the blocked signal in block units and outputting an index signal representing the index of the selected vector, , The block signal is divided into three states, that is, a block in which all pixels have a minimum pixel value, a block in which all pixels have a maximum pixel value, and another block, which is ternarized and output as a ternary pattern signal 3 The binarizing means and the three-valued pattern signal are combined into a plurality of blocks in which all pixels have a minimum pixel value and a plurality of blocks in which all pixels have a maximum pixel value. Layering means for layering and outputting a layered pattern signal for each layer, and variable-length coding the layered pattern signal and outputting it as a layered pattern variable-length coded signal, the layered pattern signal having a predetermined condition. And a layered pattern variable length coding means for outputting a coding control signal to the index variable length coding means, and the index signal based on the coding control signal. A vector index variable length coding means for analyzing the correlation to perform variable length coding and outputting as an index variable length coded signal, the hierarchical pattern variable length coded signal and the index signal based on a predetermined condition. Encode,
And a code string generating means for generating a code string and outputting it as an encoded signal.

An image decoding apparatus according to an eighth invention uses as an input signal a signal encoded by the image encoding apparatus according to the seventh invention,
A code which is separated into a hierarchical pattern decoded signal and an index decoded signal, the hierarchical pattern decoded signal is output to a ternary pattern generation means and a rotation state determination means, and the index decoded signal is output to a vector dequantization means. A column decoding means, a ternary pattern generating means for reproducing a hierarchical structure from the hierarchical pattern decoded signal and expanding the ternary pattern to output a ternary pattern decoded signal, and a vector from the index decoded signal , A vector dequantizing means for generating a decoding screen for each processing block and outputting a block decoded signal, and analyzing a pattern from the hierarchical pattern decoded signal, A rotation state determination means that analyzes the rotation state and outputs a rotation control signal, and the block decoding signal based on the rotation control signal A block rotating means for rotating the block and outputting it as a block rotation decoded signal;
Using the value pattern decoded signal and the block rotation decoded signal as an input signal, the position of the vector quantized block is determined from the ternary pattern decoded signal, and the block rotation decoded signal is assigned to the position. In the case of blocks other than the vector quantization block, the decoded signal generation in which the minimum pixel value or the maximum pixel value is assigned to all the pixels of the block according to the value of the ternary pattern decoded signal and the decoded signal is output And means.

An image coding apparatus according to a ninth aspect of the present invention is a signal representing the image transparency information, area shape information, and the like. The input signal, which is the image to be coded, is divided into blocks, and the blocked signal is obtained. As a blocking means for outputting to the ternarizing means and the vector quantizing means, the blocking signal is a block in which all pixels have a minimum pixel value, a block in which all pixels have a maximum pixel value, and other blocks. A ternarization unit that divides the case into three states and outputs the ternary pattern signal as a ternary pattern signal, and the ternary pattern signal includes a plurality of blocks in which all pixels have a minimum pixel value and all pixels have a maximum pixel value. A plurality of blocks are respectively grouped into a layer, and a layering means for outputting the layered pattern signal to the code string generation means and the layered pattern comparison means, and a codebook output by the vector quantization means. , Codebook ternarization that divides into three states, a block in which all pixels have minimum pixel values, a block in which all pixels have maximum pixel values, and other blocks, and outputs as a codebook ternary pattern signal Means, and the codebook ternary pattern signal, a plurality of blocks in which all pixels have a minimum pixel value and a plurality of blocks in which all pixels have a maximum pixel value are respectively layered together, and a codebook layered pattern signal is obtained. The codebook layering means for outputting to the layering pattern comparing means is compared with the layering pattern signal and the codebook layering pattern signal, and the correlation of the layering pattern is analyzed to obtain the vector quantization control signal. A hierarchical pattern comparison means for outputting,
Based on the vector quantization control signal, the block signal is vector-quantized in block units, the index signal representing the index of the selected vector is output to the code string generation means, and the codebook is converted into the codebook ternaryization means. The vector quantization means for outputting and the code string generating means for coding the hierarchical pattern signal and the index signal based on a predetermined condition to generate a code string and outputting as a coded signal. .

An image coding apparatus according to a tenth aspect of the present invention is a signal representing image transparency information, area shape information, etc., which is an image to be coded, and is divided into blocks for each processing unit. As a block forming means for outputting to the transparency / region shape information encoding means and the motion compensation determining means, the reference screen signal is moved based on the motion vector signal.
The motion compensating means for deforming and outputting the motion compensating screen signal is compared with the motion compensating screen signal and the blocked signal, and if the difference value is less than a predetermined value, the motion vector output control signal is selected as the motion vector output selecting means. And the motion compensation selection as an additional information signal at the same time. If the difference value is larger than a predetermined value, the coding control signal is output to the transparency / region shape information coding means, and at the same time, the intra-picture coding selection is performed. Based on the motion compensation determination means for outputting as an additional information signal, and the encoding control signal, if intra-frame encoding is selected, the blocking signal, the encoding described in claim 1 Transparency / region shape information coding means for coding based on a device and outputting as a coded signal; decoding means for decoding the coded signal and outputting the decoded signal to a memory means; When the motion compensation is selected on the basis of the motion vector output control signal, the memory device stores the encoded signal and outputs it to the motion compensation device as a reference screen signal at the time of encoding at the next time. And a motion vector output selecting means for outputting an output signal.

[0023]

In the image coding apparatus of the first invention, the area shape information and the transparency information image do not contain much intermediate level values, and most of the pixels have the maximum pixel value or the minimum pixel value.
Since it has the property that it can be expressed in binary, the block containing the intermediate level value is processed by vector quantization, and the binary information that occupies most of it is input by performing hierarchical ternary pattern coding. Efficient encoding that matches the characteristics of the image is possible.

In the image coding apparatus of the second invention,
By performing the vector quantization process of the block including the intermediate level value and the hierarchical ternary pattern coding of the binary information in parallel, the following becomes possible.

(1) Control of code amount of vector quantization based on hierarchical structure of ternary pattern. (2) Set the size of the vector quantization processing block to the hierarchical 3
Encoding of finer mid-level value patterns by making them smaller than the processing blocks of value pattern encoding.

(3) Code number of vector quantization and layering 3
Efficient coding by performing variable length coding independently on the value pattern and the nature of each occurrence probability.

In the image decoding apparatus of the third invention,
By decoding the hierarchical ternary pattern coded signal, the binary information is decoded and the position of the vector quantized block is determined, and the vector quantized codebook is decoded according to the position. Since this is good, it is possible to perform decoding that does not require the coding position information of the vector quantization block as additional information.

In the image coding apparatus of the fourth invention,
The area shape information and the transparency information image, which are the input signals,
It is determined whether it is a binary image or a multi-valued image, vector quantization is performed only in the case of a multi-valued image, vector quantization is omitted in the case of a binary image, and only hierarchical ternary pattern coding is performed. By doing so, it is possible to realize a reduction in the amount of encoding processing and at the same time, to perform efficient encoding that matches the properties of the input image.

In the image coding apparatus of the fifth invention,
The area shape information and the transparency information image, which are the input signals,
Determine the type of image, whether it is a character image, a natural image, or a computer graphics image,
By preparing a vector quantization codebook for each type of image, efficient encoding that matches the nature of the image becomes possible.

In the image coding apparatus of the sixth invention,
Efficient encoding can be performed by analyzing the frequency of appearance of the pattern for each layer of the hierarchical ternary pattern and performing variable length encoding for each layer. Furthermore, by performing variable-length coding on the pattern of the lower layer according to the pattern of the upper layer, that is, by coding using the correlation between layers, efficient coding becomes possible.

In the image coding apparatus of the seventh invention,
By variable-length coding the vector quantization code number using the hierarchical ternary pattern, efficient coding can be performed by using the correlation between the hierarchical structure and the pattern of the vector quantization block. Furthermore, according to the hierarchical ternary pattern, the vector quantization block corresponding to the hierarchical pattern is rotated to concentrate the appearance frequency on a specific vector quantization pattern, and variable length coding is performed to achieve more efficient coding. Is possible.

In the image decoding apparatus of the eighth invention,
Performing decoding corresponding to the encoding device of the seventh invention, with respect to the rotation of the vector quantization block, since the rotation can be uniquely restored if the hierarchical ternary pattern is decoded,
It is possible to perform decoding without adding information about block rotation.

In the image coding apparatus of the ninth invention,
Before performing a vector search, the vector quantized codebook vector hierarchical ternary pattern is compared with the encoded block hierarchical ternary pattern to limit the vector to be searched. Vector quantization is possible.

In the image encoding device of the tenth invention,
When encoding a moving image of transparency / region shape information, determination is made based on a motion compensation difference signal. If the difference is small, only motion compensation is performed, and if the difference is large, the code of the image encoding device. By intra-coding by the coding method, it is possible to perform efficient coding using the correlation in the time direction of moving images without preparing a vector quantization codebook for inter-coding. .

[0035]

1 is a block diagram of an image coding apparatus according to an embodiment of the first invention.

In the figure, 1001 is an input signal, 1002 is a blocking means, 1003 is a blocking signal, 1004 is a vector quantizing means, 1005 is an index signal, 1006 is a vector quantization pattern signal, and 1007 is a ternarizing means. , 1008 is a ternary pattern signal, 1009 is a layering means, 1010 is a layered pattern signal, 10
Reference numeral 11 is a code string generating means, and 1012 is a coded signal.

The operation of the image coding apparatus of the present embodiment configured as above will be described below.

The blocking means 1002 divides the input signal 1001 which is the signal to be coded into blocks which are units of coding processing, and outputs a blocked signal 1003. As a coding unit, a rectangular block of 16 pixels in the vertical direction and 16 pixels in the horizontal direction and a rectangular block of 8 pixels in the vertical direction and 8 pixels in the horizontal direction are usually selected, but the encoding unit is not limited to this.

The vector quantizer 1004 searches the codebook for the nearest vector pattern in block units of the blocked signal 1003, and the code number of the vector pattern is used as the index signal 1005 in the code string generator 1011. Then, the pattern of the vector quantization block is output to the ternarization means 1007 as a vector quantization pattern signal 1006. At this time, the input signal is an image showing the area shape information and the transparency information, and the intermediate level value is small, and most of the information is binary information. Therefore, when vector quantization is performed, the block containing the intermediate level value is included. Is.
However, vector quantization may be performed when a vector in which all pixels have the maximum pixel value or all pixels have the minimum pixel value is prepared in the codebook and the intermediate level value is not included. The vector quantization pattern signal 1006 is a signal that can identify three types of blocks: all pixels have a maximum pixel value block, all pixels have a minimum pixel value block, and other blocks. The maximum pixel value and the minimum pixel value are, for example,
When the input signal is represented by 8 bits, the maximum pixel value is 255 and the minimum pixel value is 0.

Based on the vector quantization pattern signal 1006, the ternarization means 1007 classifies the pixel into three blocks, that is, a block in which all pixels have the maximum pixel value, a block in which all pixels have the minimum pixel value, and other blocks. The three-valued pattern signal 1008
Is output to the layering means 1009. FIG. 19 shows the explanation of the ternarization of the vector quantization pattern.

In the figure, as an example, a vertical block is divided into rectangular blocks of 4 vertical pixels by 16 horizontal pixels, and vector quantization is performed in block units. Therefore, the vector quantization pattern is a block of vertical 4 horizontal 4. As shown in the figure, the pattern of this block can be classified into three types, that is, a block in which all pixels have a maximum pixel value, a block in which all pixels have a minimum pixel value, and a block other than that, and can be ternarized.

In the layering means 1009, the ternary pattern signal 100
The pattern is hierarchized based on 8, and the hierarchized pattern signal 1010 is output to the code string generation means 1011.

Various methods can be selected for the hierarchization, such as 2-branch tree hierarchization, 4-branch tree hierarchization, and multi-branch tree hierarchization. Description will be given based on hierarchization. FIG. 20 shows an explanatory diagram regarding codeword generation of a ternary pattern. This figure is a codeword assuming quadtree hierarchization, and a codeword is generated based on a pattern of vertical 2 horizontal 2 patterns. As shown in the figure, 0 (BLACK) for the minimum pixel value for all pixels and 2 (WH for the maximum pixel value for all pixels.
ITE), otherwise 1 (GRAY), 4 blocks together,
Generate a codeword as a 3-bit value. In the case of the illustrated pattern, 14 code words are assigned.

FIG. 21 shows an explanatory diagram regarding the hierarchization of the ternary pattern. The upper layer is generated by grouping four blocks, and the generation conditions are 0 (BLACK) for the upper layer block if all 4 are BLACK, 2 (WHITE) for the upper layer block if all 4 are WHITE, Otherwise, set to 1 (GRAY). A hierarchical structure is formed by repeating this process.

The code string generating means 1011 encodes the layered pattern signal 1010 and the index signal 1005, and outputs the coded signal 1012. As shown in FIG. 21, if the upper layer pattern is encoded first, the 0 can be omitted when the lower layer contains 0, and by omitting it, efficient encoding can be realized.

As described above, according to the present embodiment, the area shape information and the transparency information image do not contain much intermediate level values, and most of the pixels can be represented by the binary value of the maximum pixel value or the minimum pixel value. Of the processing unit blocks,
Blocks containing intermediate level values are vector quantized,
By processing the vector quantization pattern by the hierarchical ternary pattern coding, efficient coding that matches the property of the input image becomes possible.

FIG. 2 shows a block diagram of an image coding apparatus in the first embodiment of the second invention. In the figure, 1001 is an input signal, 1002 is a blocking means, 1003 is a blocking signal,
1014 is a vector quantization means, 1015 is an index signal,
1017 is a ternarization means, 1018 is a ternary pattern signal, 1019 is a hierarchization means, 1020 is a hierarchization pattern signal, 1021 is a code string generation means, and 1022 is an encoded signal.

The operation of the image coding apparatus of the present embodiment configured as described above will be described below.

The blocking means 1002 divides the input signal 1001 which is the signal to be coded into blocks which are units of coding processing, and outputs the blocked signal 1003 to the ternarization means 1017 and the vector quantization means 1014. As a coding unit,
Usually, a rectangular block of 16 pixels in the vertical direction and 16 pixels in the horizontal direction and a rectangular block of 8 pixels in the vertical direction and 8 pixels in the horizontal direction are often selected, but not limited to this.

The vector quantizing means 1014 retrieves the closest vector pattern from the codebook in block units of the blocked signal 1003, and the code number of the vector pattern is given to the code string generating means 1011 as the index signal 1015. Output. At this time, the input signal is an image showing the area shape information and the transparency information, and the intermediate level value is small, and most of the information is binary information. Therefore, when vector quantization is performed, the block containing the intermediate level value is included. Is. However, vector quantization may be performed when a vector in which all pixels have the maximum pixel value or all pixels have the minimum pixel value is prepared in the codebook and the intermediate level value is not included.

The ternarization means 1017 uses the blocked signal 1003.
Is classified into three blocks, that is, a block in which all pixels have a maximum pixel value, a block in which all pixels have a minimum pixel value, and a block other than that, and is ternarized, and a ternary pattern signal 1018 is output to the layering means 1019. The method of ternarization is the same as in FIG. As shown in the figure, the block pattern can be classified into three types, that is, a block in which all pixels have a maximum pixel value, a block in which all pixels have a minimum pixel value, and a block other than that, and can be ternarized.

In the layering means 1019, the ternary pattern signal 101
The pattern is hierarchized based on 8, and the hierarchized pattern signal 1020 is output to the code string generation means 1021. The hierarchization can be performed by the same method as that shown in FIG. 20 described in the first invention.

The code string generating means 1021 encodes the layered pattern signal 1020 and the index signal 1015, and outputs the encoded signal 1022. As shown in FIG. 21, when the pattern of the upper layer is already coded and 0 is included in the lower layer, 0 can be omitted. By omitting 0, efficient coding can be realized.

As described above, according to the present embodiment, the area shape information and the transparency information image do not contain much intermediate level values, and most of the pixels can be represented by two values of the maximum pixel value or the minimum pixel value. Of the processing unit blocks,
Blocks containing intermediate level values are vector quantized,
On the other hand, the binary information that occupies most of them can be efficiently coded according to the property of the input image by performing the hierarchical ternary pattern coding.

The difference between the first invention and the present invention is that the vector quantization process of the block containing the intermediate level value and the hierarchical ternary pattern coding of the binary information are performed in parallel, and It becomes possible as in the embodiment shown.

(1) Control of code amount of vector quantization based on hierarchical structure of ternary pattern (second embodiment).

(2) Encoding of a more precise intermediate level value pattern by making the size of the processing block of vector quantization smaller than that of the processing block of hierarchical ternary pattern encoding (third embodiment) ).

(3) Code number of vector quantization and layering 3
Efficient coding by performing variable length coding independently on the value pattern and the nature of each occurrence probability.
Example).

FIG. 3 shows a block diagram of an image coding apparatus according to the second embodiment of the second invention. The present embodiment has a configuration in which a code amount control function is added to the configuration of the first embodiment, and has a configuration in which a code amount calculation means 1023 and a vector quantization control signal 1024 in the figure are added. ing.

With respect to the operation of the image coding apparatus of the present embodiment configured as described above, the description of the configuration common to that of the first embodiment will be omitted, and the portion related to the added code amount control function will be described. Only described below.

The layering means 1019 is used for layering pattern signal 1020.
Is output to the code string generation means 1021 and the code amount calculation means 1023. In the code amount calculation means 1023, the hierarchical pattern signal 1020
, The vector quantization control signal 1024 is output to the vector quantization means 1014 to reduce the vector quantization, and control of the entire code amount can be realized. . As an example of controlling the overall code amount, if the code amount of the layered pattern is equal to or larger than a predetermined threshold value, the number of vector quantization codebooks that are candidates is limited to set the generated code amount of vector quantization. A method of controlling by reducing or stopping vector quantization can be considered. Further, it is also possible to reflect the calculation result of the code amount in the encoding of the layered pattern and to control the code amount such that the layers below a certain level are cut off to reduce the code amount.

As described above, according to the present embodiment, by performing the vector quantization and the hierarchical ternary pattern coding in parallel, one coding controls the other coding, It is possible to control the entire code amount. In particular, when the band of the transmission path of the encoded signal is limited or the capacity of the storage device is limited, such a code amount control function plays a large role.

FIG. 4 shows a block diagram of an image coding apparatus in the third embodiment of the second invention. The present embodiment has a structure in which a small block forming function of a vector quantization block is added to the structure of the first embodiment, and a small block forming unit 1031 and a small block forming signal 1032 in the figure are added. It is configured.

With respect to the operation of the image coding apparatus of the present embodiment configured as described above, description of the configuration common to that of the first embodiment is omitted, and the added vector quantization block is divided into small blocks. Only the part related to the function will be described below.

The small block forming means 1031 is a block forming signal.
The 1003 is divided into smaller blocks and output as a small block signal 1032 to the vector quantization means 1014. Generally, in vector quantization, the size of the codebook is limited. Therefore, when the block for vector quantization becomes large, the number of indexes of the candidate codebook decreases, and the coding that reproduces a fine pattern is performed. Becomes impossible. If you make the vector quantization block small,
As the number of codebook indexes increases accordingly,
It means that finer patterns can be encoded. That is, the size of the vector quantization processing block
By making it smaller than the processing block of ternary pattern coding, a finer pattern of intermediate level values can be realized by vector quantization by a small block.

As described above, according to this embodiment, the vector quantization and the hierarchical ternary pattern coding are executed in parallel, and the size of the vector quantization block is reduced, so that the code of the same capacity can be obtained. When a book is used, it is possible to perform encoding that can reproduce a finer distribution pattern of intermediate level values.

FIG. 5 shows a block diagram of an image coding apparatus in the fourth embodiment of the second invention. The present embodiment has a configuration in which a variable length coding function for a hierarchical pattern and an index is added to the configuration of the first embodiment. Hierarchical pattern variable length coding means 1041 in the figure, hierarchical The pattern variable length coded signal 1042, the index variable length coding means 1044, the index variable length coded signal 1045, the code string generation means 1043, and the coded signal 1046 are changed.

With respect to the operation of the image coding apparatus of the present embodiment configured as described above, description of the configuration common to that of the first embodiment will be omitted, and the layered pattern and the added configuration will be omitted. Only the part related to the variable length coding function of the index will be described below.

The layered pattern variable length coding means 1041 calculates the frequency of appearance of the pattern from the layered pattern signal 1020 to perform variable length coding, and sends it to the code string generation means 1043 as the layered pattern variable length coded signal 1042. Output. Fig. 22
An explanatory diagram regarding the appearance frequency of the ternary pattern is shown in FIG. Although the figure is an example, as in (a) of the figure, the distribution of 0 (BLACK), which represents a block in which all pixels have the minimum pixel value, and 2 (WHITE), which represents a block in which all pixels have the maximum pixel value, The biased pattern has a high appearance frequency, and the checkerboard pattern as illustrated in FIG. 7B has a low appearance frequency. That is, it is possible to improve the coding efficiency by calculating such an appearance frequency and assigning a code having a short code length to a pattern having a high appearance frequency.

On the other hand, index variable length coding means 1044
Then, the index signal 1015 is subjected to variable length coding by calculating the appearance frequency of the code number, and the index variable length coded signal 1045 is output to the code string generation means 1043. The code string generation means 1043 generates a code string by combining the hierarchical pattern variable length coded signal 1042 and the index variable length coded signal 1045, and outputs it as a coded signal 1046.

As described above, according to this embodiment, the vector quantization and the hierarchical ternary pattern coding are executed in parallel, the frequency of appearance of the pattern is calculated for each, and the variable length coding is performed. By this, efficient encoding is possible.

FIG. 6 shows a block diagram of an image decoding apparatus in an embodiment of the third invention. In the figure, 2001 is an input signal, 2002 is a code string decoding means, 2003 is an index decoding signal, 2004 is a vector dequantization means, 2005 is a block decoding signal, 2006 is a hierarchical pattern decoding signal, and 2007 is 3
Value pattern generation means, 2008 is a ternary pattern decoded signal, 2
Reference numeral 009 is a decoded signal generation means, and reference numeral 2010 is a decoded signal.

The operation of the image decoding apparatus of the present embodiment configured as above will be described below.

In the code string decoding means 2002, the first and second
The input signal 2001, which is a signal encoded by the encoding device of the invention of claim 1, is separated into a hierarchical pattern decoded signal 2006 and an index decoded signal 2003, which are respectively given to a ternary pattern generation means 2007 and a vector dequantization means 2004. Output.

In the ternary pattern generation means 2007, it was omitted at the time of encoding based on the hierarchical pattern decoded signal 2006.
By substituting 0, the hierarchical structure is restored to generate a ternary pattern of the lowest hierarchy, and the ternary pattern decoded signal 2008 is output to the decoded signal generation means 2009. The pattern generated here represents, so to speak, the position information of the vector quantization block.

The vector dequantization means 2004 selects the distribution pattern of the intermediate level values corresponding to the number of the index decoded signal 2003 from the codebook and outputs it as the block decoded signal 2005 to the decoded signal generation means 2009. To do.

The decoded signal generation means 2009 determines the position of the vector-quantized block based on the ternary pattern decoded signal 2008, and sets the block decoded signal 20 at that position.
Substitute 05 for decryption.

Further, at the positions of the pattern in which all pixels represent the minimum pixel value block and all pixels represent the maximum pixel value block in the ternary pattern, the minimum pixel value and the maximum pixel value are substituted into each block. The signal decoded in this way is output as a decoded signal 2010.

As described above, according to this embodiment, hierarchization
By decoding the ternary pattern coded signal, the decoding of the binary information and the position of the vector quantized block can be determined, and the vector quantized codebook can be decoded according to the position. Decoding that does not require the coding position information of the vector quantization block as additional information is possible.

FIG. 7 shows a block diagram of an image coding apparatus according to an embodiment of the fourth invention. In the figure, 1001 is an input signal, 1002 is a blocking means, 1003 is a blocking signal, 10
Reference numeral 51 is a binary / multi-value determination means, 1052 is a selection control signal, 1053, 1
054 is a selection unit, 1055 is a selected block signal, 1104 is a vector quantization unit, 1105 is an index signal, 1106 is a vector quantization pattern signal, 1107 is a ternarization unit, 1108 is a ternary pattern signal, and 1109 is a layering unit. , 1110 is a layered pattern signal, 1111 is a code string generating means, and 1112 is a coded signal.

The operation of the image coding apparatus of the present embodiment configured as above will be described below.

The blocking means 1002 divides the input signal 1001 which is the signal to be coded into blocks which are units of coding processing, and the blocked signal 1003 is selected by the selecting means 1053 and binary / binary.
It outputs to the multi-value determination means 1051. As a coding unit, a rectangular block of 16 pixels in the vertical direction and 16 pixels in the horizontal direction and a rectangular block of 8 pixels in the vertical direction and 8 pixels in the horizontal direction are usually selected, but the encoding unit is not limited to this.

The binary / multi-value determination means 1051 analyzes the blocked signal 1003, and if multi-value is included, vector quantization is performed, and if it is only binary, vector quantization is not performed and layering is performed. The selection control signal 1052 is output to the selection means 1053 and the selection means 1054 so that only the value pattern coding is performed.
As for the binary / multivalued determination method, for example, the frequency distribution of pixel values is taken, and when only binary values appear, binary values are used, and other values are multivalued. Even if the distribution is three or more, if the intermediate level value is extremely small, it will be forced.
It may be binarized and processed.

In the selection means 1053, when vector quantization is performed on the basis of the selection control signal 1052, a block signal is generated.
1003 is output to the vector quantization means 1104, and when the vector quantization is not performed, the blocked signal 1003 is selected by the selection means 10
Output to 54.

The vector quantizing means 1104 searches the codebook for the nearest vector pattern in block units of the blocked signal 1003, and the code number of the vector pattern is given to the code string generating means 1111 as the index signal 1105. Output and vector quantization pattern signal 11
06 is output to the selection means 1054.

In the selecting means 1054, when the vector quantization is performed based on the selection control signal 1052, the vector quantization pattern signal 1106 is set as the selected block signal 1055.
When the signal is output to the binarizing means 1107 and the vector quantization is not performed, the blocked signal 1003 is selected and the blocked signal 1055 is selected.
Is output to the ternarization means 1107.

In the ternarization means 1107, the selected block signal 1
055 is categorized into three blocks, that is, a block in which all pixels have a maximum pixel value, a block in which all pixels have a minimum pixel value, and a block other than that, and ternarization is performed, and a ternary pattern signal 1108 is made into a layering means 1109
Output to. The method of ternarization is the same as in Fig. 19, and as shown in the figure, all pixels are classified into three blocks: blocks with maximum pixel value, blocks with minimum pixel value for all pixels, and other blocks. It can be quantified.

In the layering means 1109, the ternary pattern signal 110
The pattern is hierarchized based on 8, and the hierarchized pattern signal 1110 is output to the code string generation means 1111. The hierarchization can be performed by the same method as that shown in FIG. 20 described in the first invention.

The code string generating means 1111 encodes the layered pattern signal 1110 and the index signal 1105 if vector quantization is performed, and if the vector quantization is not performed, layered pattern The signal 1110 is encoded and output as an encoded signal 1112. As shown in FIG. 21, when the upper layer pattern is already encoded and the lower layer contains 0, 0 can be omitted, and by omitting 0, efficient encoding can be realized. .

As described above, according to this embodiment, it is determined whether the area shape information and the transparency information image which are the input signals are the binary image or the multivalued image. Only the vector quantization is performed, in the case of a binary image, the vector quantization is omitted, and only the hierarchical ternary pattern coding is performed, thereby reducing the coding processing amount and at the same time,
It is possible to perform efficient encoding that matches the properties of the input image.

FIG. 8 shows a block diagram of an image coding apparatus according to an embodiment of the fifth invention. In the figure, 1001 is an input signal, 1002 is a blocking means, 1003 is a blocking signal, 10
61 is an image determination means, 1062 is a codebook selection control signal,
1063 is the codebook 1 signal, 1064 is the codebook 2 signal,
1065 is a codebook selection means, 1066 is a selected codebook, 1204 is a vector quantization means, 1205 is an index signal, 1206 is a vector quantization pattern signal, 1007 is a ternarization means, 1008 is a ternary pattern signal, and 1009 is a hierarchy. Means, 101
0 is a hierarchical pattern signal, 1211 is a code string generating means, 1212
Is a coded signal.

The operation of the image coding apparatus of the present embodiment configured as above will be described below.

The blocking means 1002 divides the input signal 1001 which is the signal to be coded into blocks which are units of encoding processing, and outputs the blocked signal 1003 to the image determination means 1061 and the vector quantization means 1204. As a coding unit, a rectangular block of 16 pixels in the vertical direction and 16 pixels in the horizontal direction and a rectangular block of 8 pixels in the vertical direction and 8 pixels in the horizontal direction are usually selected, but the encoding unit is not limited to this.

In the image judging means 1061, the blocking signal 10
Based on 03, the image is analyzed to determine whether the input image is the character region shape and transparency information, the natural image region shape information and transparency information, or the computer graphics region shape and transparency information. It is determined whether or not there is, and the codebook selection control signal 1062 is output to the codebook selection means 1065 according to the type of image. Note that if the type of input image is determined in advance, it may be switched manually. As for the image determination method, the statistical distribution such as frequency distribution of pixel values, average value, and variance value is calculated,
A feature amount corresponding to each image type may be determined in advance as a criterion and the input images may be classified based on the criterion.

In the codebook selection means 1065, the codebook corresponding to the type of image is codebook 1 signal 1063 and codebook 2 based on the codebook selection control signal 1062.
Select from signal 1064 and select codebook 1066,
Output to the vector quantizer 1204. Although two codebooks are prepared here, the number of codebooks is not limited to two. The codebook may be prepared according to the above-mentioned image classification, but at this time, for example, the image classification may be binary or multivalued and the codebook may be prepared for each.

The vector quantizer 1204 searches the selected codebook 1066 for the closest vector pattern in block units of the blocked signal 1003, and generates a code string as the index signal 1205 using the code number of the vector pattern. It outputs to the means 1211 and outputs the pattern of the vector quantization block to the ternarization means 1007 as the vector quantization pattern signal 1206.

Based on the vector quantization pattern signal 1206, the ternarization means 1007 classifies the image into three blocks, that is, a block in which all pixels have the maximum pixel value, a block in which all pixels have the minimum pixel value, and other blocks. The three-valued pattern signal 1008
Is output to the layering means 1009. The ternarization of the vector quantization pattern is the same as in FIG. As shown in the figure, all pixels can be classified into three types, that is, a block having a maximum pixel value, a block having a minimum pixel value for all pixels, and a block other than that.

In the layering means 1009, the ternary pattern signal 100
The pattern is hierarchized based on 8, and the hierarchized pattern signal 1010 is output to the code string generation means 1011. For the layering, the layering method shown in FIG. 21 can be used.

The code string generating means 1011 encodes the layered pattern signal 1010 and the index signal 1005, and outputs the coded signal 1012. As shown in FIG. 21, when the pattern of the upper layer is known and 0 is included in the lower layer, 0 can be omitted, and by omitting it, efficient compression can be realized.

As described above, according to this embodiment, whether the region shape information and the transparency information image which are the input signals are character images, natural images, computer graphics images, By determining the type and preparing a vector quantization codebook for each type of image, efficient encoding that matches the nature of the image becomes possible.

FIG. 9 shows a block diagram of an image coding apparatus in the first embodiment of the sixth invention. In the figure, 1001 is an input signal, 1002 is a blocking means, 1003 is a blocking signal,
1014 is a vector quantization means, 1015 is an index signal,
1017 is a ternary conversion means, 1018 is a ternary pattern signal, 1071 is a layering means, 1072 is a layer 1 pattern signal, 1073 is a layer 1 pattern variable length coding means, 1074 is a layer 1 pattern variable length coding signal, 1075 Is a layer 2 pattern signal, 1076 is a layer 2 pattern variable length coding means, 1077 is a layer 2 pattern variable length coding signal, 1078 is a code string generating means, and 1079 is a coded signal.

The operation of the image coding apparatus of the present embodiment configured as above will be described below.

The blocking means 1002 divides the input signal 1001 which is the signal to be coded into blocks which serve as an encoding processing unit, and outputs the blocked signal 1003 to the ternarization means 1017 and the vector quantization means 1014. As a coding unit,
Usually, a rectangular block of 16 pixels in the vertical direction and 16 pixels in the horizontal direction and a rectangular block of 8 pixels in the vertical direction and 8 pixels in the horizontal direction are often selected, but not limited to this.

The vector quantizing means 1014 searches the codebook for the nearest vector pattern in block units of the blocked signal 1003, and the code number of the vector pattern is given to the code string generating means 1011 as the index signal 1015. Output.

The ternarization means 1017 uses the blocked signal 1003.
Is categorized into a block in which all pixels have the maximum pixel value, a block in which all pixels have the minimum pixel value, and a block other than that, and is ternarized, and a ternary pattern signal 1018 is output to the layering means 1009. The method of ternarization is the same as in FIG. In the figure, as an example, 16 pixels in the vertical direction and 4 pixels in the horizontal direction out of 16 pixels in the horizontal direction
It is divided into rectangular blocks of pixels and vector quantization is performed in block units. Therefore, the vector quantization pattern is a block of vertical 4 horizontal 4. As shown in the figure, the pattern of this block can be classified into three types, that is, a block in which all pixels have a maximum pixel value, a block in which all pixels have a minimum pixel value, and a block other than that, and can be ternarized.

In the layering means 1019, the ternary pattern signal 101
Patterns are hierarchized based on 8 and each layer is layer 1
The pattern signal 1072 is converted into the layer 1 pattern variable length coding means 107.
3, the layer 2 pattern signal 1075 is output to the layer 2 pattern variable length coding means 1076, respectively. Regarding layering, first
The invention can also be carried out by the same method as that shown in FIG. In addition, although the case where the number of layers is two has been described in the present embodiment, the number of layers is not limited to two.

The layer 1 pattern variable length coding means 1073 calculates the appearance frequency of the pattern in the layer 1 from the layer 1 pattern signal 1072 and performs variable length coding, and generates a code string as a layer 1 pattern variable length coded signal 1074. Output to the means 1078. The higher the hierarchy is, the more the GRAY appears in FIG. 20, and the pattern frequency distribution is different for each hierarchy. Therefore, it is possible to improve the coding efficiency by calculating the appearance frequency for each layer and assigning a code having a short code length to a pattern having a high appearance frequency.

Similarly, layer 2 pattern variable length coding means 1
In 076, from the layer 2 pattern signal 1075, the frequency of appearance of the pattern in layer 2 is calculated and subjected to variable length coding, and layer 2
Code string generating means 10 as pattern variable length coded signal 1077
Output to 78.

The code string generating means 1078 encodes the layer 1 pattern variable length coded signal 1074, the layer 2 pattern variable length coded signal 1077 and the index signal 1015, and outputs them as a coded signal 1079. As shown in FIG. 21, when the pattern of the upper layer is known and 0 is included in the lower layer, 0 can be omitted, and by omitting 0, efficient encoding can be realized.

As described above, according to this embodiment, hierarchization is performed.
Efficient encoding is possible by analyzing the frequency of appearance of patterns for each layer of the ternary pattern and performing variable length encoding for each layer.

FIG. 10 shows a block diagram of an image coding apparatus in the second embodiment of the sixth invention. The present invention has a configuration in which a coding control function between layers is added to the first embodiment, and a coding control signal 1080 in the drawing is added.

With respect to the operation of the image coding apparatus of the present embodiment configured as described above, description of the configuration common to that of the first embodiment is omitted, and the added configuration is variable between layers. Only the part related to the function of the long coding control will be described below.

The layer 1 pattern variable length coding means 1073 is
The coding control signal 1080 for controlling the coding of the layer 2 pattern variable length coding means 1076 is output. Here, for example, assuming that the layer 1 is the upper layer with respect to the layer 2, the appearance frequencies of the patterns of the lower layer are classified and calculated based on the pattern of the upper layer, and the variable length coding is performed for each pattern of the upper layer. By doing so, it is possible to further improve the coding efficiency.

Also, by using the correlation with the surrounding blocks in the upper layer, it is possible to concentrate the appearance frequency and shorten the average code length of the variable length coding. FIG. 23 (a) shows an explanatory diagram regarding the appearance frequency of the ternary pattern by the correlation with the surrounding blocks. As shown in the figure, if the upper, left, and licked values in the upper layer are BLACK, there is a high probability that BLACK will appear in the upper left block in the current layer. In the figure, the frequency of occurrence of pattern A is higher than that of B. Considered to be many.

As described above, according to this embodiment, hierarchization is performed.
For each tier of the ternary pattern, analyze the appearance frequency of the pattern, further variable length coding the pattern of the lower layer according to the pattern of the upper layer, that is, by using the correlation between the layers, Efficient encoding is possible.

FIG. 11 shows a block diagram of an image coding apparatus in the first embodiment of the seventh invention. In the figure, 1001 is an input signal, 1002 is a blocking means, 1003 is a blocking signal,
1014 is a vector quantization means, 1015 is an index signal,
1017 is a ternarization means, 1018 is a ternary pattern signal, 1019 is a hierarchization means, 1020 is a hierarchization pattern signal, 1081 is an index variable length coding means, 1082 is an index variable length coding signal, and 1083 is a hierarchization pattern. Variable length coding means, 1084 is a hierarchical pattern variable length coded signal, 1085 is a coding control signal, 1086 is a code string generation means, and 1087 is a coded signal.

The operation of the image coding apparatus of the present embodiment configured as described above will be described below.

The blocking means 1002 divides the input signal 1001 which is the signal to be coded into blocks which are units of coding processing, and outputs the blocked signal 1003 to the ternarization means 1017 and the vector quantization means 1014. As a coding unit,
Usually, a rectangular block of 16 pixels in the vertical direction and 16 pixels in the horizontal direction and a rectangular block of 8 pixels in the vertical direction and 8 pixels in the horizontal direction are often selected, but not limited to this.

The ternarization means 1017 uses the blocked signal 1003.
Is categorized into a block in which all pixels have the maximum pixel value, a block in which all pixels have the minimum pixel value, and a block other than that, and is ternarized, and a ternary pattern signal 1018 is output to the layering means 1009. The method of ternarization is the same as in FIG. In the figure, as an example, 16 pixels in the vertical direction and 4 pixels in the horizontal direction out of 16 pixels in the horizontal direction
It is divided into rectangular blocks of pixels and vector quantization is performed in block units. Therefore, the vector quantization pattern is a block of vertical 4 horizontal 4. As shown in the figure, the pattern of this block can be classified into three types, that is, a block in which all pixels have a maximum pixel value, a block in which all pixels have a minimum pixel value, and a block other than that, and can be ternarized.

In the layering means 1019, the ternary pattern signal 101
Based on 8, the pattern is hierarchized and the hierarchized pattern signal 1020 is output to the hierarchized pattern variable length coding means 1083. Regarding the hierarchization, FIG. 20 described in the first invention is also used.
It is possible to carry out by the method similar to.

The layered pattern variable length coding means 1083 calculates the frequency of appearance of the pattern from the layered pattern signal 1020, performs variable length coding, and sends it to the code string generation means 1086 as the layered pattern variable length coded signal 1084. Output. By thus calculating the appearance frequency and assigning a code having a short code length to a pattern having a high appearance frequency, it is possible to improve the coding efficiency. Further, the layered pattern variable length coding means 1083 outputs the coding control signal 1085 for controlling the coding of the vector quantization code number to the index variable length coding means 1081 based on the layered pattern.

The vector quantizing means 1014 retrieves the closest vector pattern from the codebook in block units of the blocked signal 1003, and the code number of the vector pattern is used as the index signal 1015 in the index variable length coding means. Output to 1081. In the index variable length coding means 1081, the index signal 1015 is
On the basis of the coding control signal 1085, after classifying by the higher hierarchical pattern of the part corresponding to the vector quantization block position, the appearance frequency of the code number is calculated and variable length coded, and the index variable length coded signal 1082 is calculated. It outputs to the code string generation means 1086.

Further, it is possible to shorten the average code length of the variable length coding by concentrating the appearance frequencies by using the correlation with the surrounding blocks in the upper hierarchy. FIG. 23 (b) shows an explanatory diagram regarding the appearance frequency of the vector quantization code pattern due to the correlation with the surrounding blocks. As shown in the figure, when the upper, left, and licked values in the upper layer are BLACK, the probability that the minimum pixel value appears in the upper left pixel in the vector quantization block is high, and the frequency of the pattern A in the figure appears. Is B
Considered more than.

In the code string generation means 1086, the hierarchical pattern variable length coded signal 1084 and the index variable length coded signal
A code string is generated by combining with 1082 and is output as a coded signal 1087.

As described above, according to this embodiment, hierarchization is performed.
By performing variable length coding of the vector quantization code number using a ternary pattern, efficient coding using the correlation between the hierarchical structure and the vector quantization block pattern becomes possible.

FIG. 12 shows a block diagram of an image coding apparatus in the second embodiment of the seventh invention. This embodiment has a configuration in which a rotation function of a vector quantization block is added to the configuration of the first embodiment, and the added configuration is shown in the figure.
1301 is a vector rotation means, 1302 is a rotation block signal, 13
03 is a rotation control signal.

With respect to the operation of the image coding apparatus of the present embodiment configured as described above, the description of the configuration common to the first embodiment will be omitted, and the rotation function of the added vector quantization block will be described. Only the relevant parts will be described below.

In the block rotation means 1301, the rotation control signal output by the layered pattern variable length coding means 1083.
Based on 1303, the blocked signal 1003 is rotated and output as a rotated block signal 1302 to the vector quantization means 1014. That is, by utilizing the directional correlation between the hierarchical pattern and the vector quantization pattern, the coded block is rotated and then the vector quantization is performed. FIG. 24 shows an explanatory diagram regarding variable length coding using correlation of directionality between layers. Here, as an example, the upper layer pattern is classified into four states of A, B, C, and D by rotating every 90 degrees. Except for special cases, if four states normally exist with equal probability, for example, for blocks belonging to three states B, C, and D, 90 degrees, 180 degrees, and 270 degrees, respectively.
It is possible to bring all the states to A by rotating them in the direction opposite to the clockwise direction. As shown in the figure, for example, the encoded block belonging to the state C is moved in the counterclockwise direction.
It can be changed to state A by rotating 180 degrees. In this way, if it is possible to concentrate on one state, the appearance frequency of a pattern in variable-length coding is determined, and the distribution concentrates on the patterns belonging to the state of A, so that the average code length of variable-length coding is shortened. be able to. At this time, the rotation state may be classified by using the upper layer of the layered pattern.

As an example of the classification method, FIG. 25 shows an explanatory diagram regarding a method of determining the directional state in the upper hierarchy. Considering the numerical values assigned to A, B, C, and D in the figure as weighting, X is the upper layer pattern of the encoded block.
After multiplying with 1, X2, X3, X4, compare the total,
The largest case is determined as the directional state.

As described above, according to this embodiment, hierarchization is performed.
It is possible to perform more efficient coding by rotating the vector quantization block according to the ternary pattern, concentrating the appearance frequency in a specific vector quantization pattern, and performing variable length coding.

FIG. 13 shows a block diagram of an image decoding apparatus in an embodiment of the eighth invention. In the figure, 2101 is an input signal, 2102 is a code string decoding means, 2103 is an index decoding signal, 2104 is a vector dequantization means, 2105 is a block decoding signal, 2106 is a block rotation means, 2107 block rotation decoding 2113 is a rotation state determination means, 2115 is a ternary pattern generation means, 2116
Is a ternary pattern decoded signal, 2117 is a decoded signal generating means, and 2118 is a decoded signal.

The operation of the image decoding apparatus of the present embodiment configured as above will be described below.

In the code string decoding means 2102, the input signal 2101 which is the signal coded by the coding device shown in the second embodiment of the seventh invention is converted into the hierarchical pattern decoding signal.
2106 and the index decoded signal 2103 are separated, and the hierarchical pattern decoded signal 2106 is sent to the ternary pattern generation means 2007 and the rotation state determination means 2111 to the index decoded signal 2103.
Is output to vector dequantization means 2104.

The rotation state determining means 2111 determines the rotation state of the vector quantization block based on the hierarchical pattern decoding signal 2106, and outputs the rotation control signal 2112 to the block rotating means.
Output to 2113. To determine the rotation state, for example, invention 7
The method shown in FIG. 25 described in the second embodiment of FIG.

In the ternary pattern generation means 2107, it was omitted at the time of encoding based on the hierarchical pattern decoded signal 2106.
By substituting 0, the hierarchical structure is restored to generate a ternary pattern of the lowest hierarchy, and the ternary pattern decoded signal 2108 is output to the decoded signal generation means 2109. The pattern generated here represents, so to speak, the position information of the vector quantization block.

The vector dequantization means 2104 selects the distribution pattern of the intermediate level values corresponding to the number of the index decoded signal 2103 from the codebook and outputs it as the block decoded signal 2105 to the block rotation means 2113.

In the block rotation means 2113, the rotation control signal
Based on 2112, the block decoded signal 2105 is rotated to generate a decoded signal generating means 2109 as a block rotated decoded signal 2114.
Output to.

The decoded signal generating means 2109 determines the position of the vector-quantized block based on the ternary pattern decoded signal 2108, and substitutes the block rotation decoded signal 2114 at that position for decoding. In addition, the minimum pixel value and the maximum pixel value are assigned to the blocks at the positions of the pattern in which all the pixels represent the minimum pixel value block and all the pixels represent the maximum pixel value block in the ternary pattern. The signal decoded in this way is output as a decoded signal 2010.

As described above, according to this embodiment, hierarchization
By decoding the ternary pattern coded signal, the decoding of the binary information and the determination of the position of the vector quantized block and the state of rotation are determined, the rotation state is restored, and then the decoding position is restored. Since it is only necessary to decode the vector quantization codebook in accordance with the above, it is possible to perform decoding that does not require the coding position information of the vector quantization block and the rotation state of the block as additional information.

FIG. 14 shows a block diagram of an image coding apparatus according to an embodiment of the ninth invention. In the figure, 1001 is an input signal, 1002 is blocking means, 1003 is a blocking signal, 14
01 is the vector quantizer, 1402 is the index signal, 10
17 is a ternarization means, 1018 is a ternary pattern signal, 1419 is a hierarchical means, 1403 is a codebook, 1404 is a codebook ternary means, 1405 codebook ternary pattern signal, 1406 is a codebook tiering means, 1407 is a codebook hierarchical pattern signal, 1408 is a hierarchical pattern comparison means, 1409 is a vector quantization control signal, 1420 is a hierarchical pattern signal, 1421 is a code string generation means, and 1422 is a coded signal.

The operation of the image coding apparatus of the present embodiment configured as above will be described below.

The blocking means 1002 divides the input signal 1001 which is the signal to be coded into blocks which are units of encoding processing, and outputs the blocked signal 1003 to the ternarization means 1017 and the vector quantization means 1014. As a coding unit,
Usually, a rectangular block of 16 pixels in the vertical direction and 16 pixels in the horizontal direction and a rectangular block of 8 pixels in the vertical direction and 8 pixels in the horizontal direction are often selected, but not limited to this.

The ternarization means 1017 uses the blocked signal 1003.
Is categorized into a block in which all pixels have the maximum pixel value, a block in which all pixels have the minimum pixel value, and a block other than that, and is ternarized, and a ternary pattern signal 1018 is output to the layering means 1009. The method of ternarization is the same as in FIG. In the figure, as an example, 16 pixels in the vertical direction and 4 pixels in the horizontal direction out of 16 pixels in the horizontal direction
It is divided into rectangular blocks of pixels and vector quantization is performed in block units. Therefore, the vector quantization pattern is a block of vertical 4 horizontal 4. As shown in the figure, the pattern of this block can be classified into three types, that is, a block in which all pixels have a maximum pixel value, a block in which all pixels have a minimum pixel value, and a block other than that, and can be ternarized.

In the layering means 1019, the ternary pattern signal 101
Based on 8, the pattern is hierarchized, and the hierarchized pattern signal 1020 is output to the hierarchized pattern comparison means 1408 and the code string generation means 1021. The hierarchization can be performed by the same method as that shown in FIG. 20 described in the first invention. Next, a hierarchical ternary pattern is obtained for the codebook used for vector quantization.

The vector quantizing means 1014 is a codebook.
1403 is output to the codebook ternarization means 1404. The codebook ternarization means 1404 classifies the codebook 1403 into three blocks, that is, a block in which all pixels have a maximum pixel value, a block in which all pixels have a minimum pixel value, and a block other than that. The ternary pattern signal 1405 is output to the codebook layering means 1406. Codebook layering means 14
In 06, based on the codebook ternary pattern signal 1405,
The patterns are hierarchized, and the codebook hierarchized pattern signal 1407 is output to the hierarchized pattern comparison means 1408.

The hierarchical pattern comparing means 1408 has a hierarchical pattern signal 1020 and a codebook hierarchical pattern signal 14
If 07 is compared and it is determined that the signals match or are substantially equal, the vector quantization control signal 1409 is output to the vector quantization means 1014. FIG. 26 shows an explanatory diagram relating to a method for speeding up vector search using a hierarchical ternary pattern.
Even in the figure, it is only necessary to search for code vectors 1 and 1 in the vector quantization codebook by comparing hierarchical ternary patterns, and searches for code vectors 2 to 4 and similar can be omitted. Is. In this way, by comparing the hierarchical pattern of the codebook and the hierarchical pattern of the encoded block, the vector to be searched by the vector quantizing means 1014 is limited, the processing is reduced, and high-speed search is realized. It becomes possible to do.

The vector quantizing means 1014 searches the codebook whose search range is limited based on the vector quantizing control signal 1409 for a vector pattern closest to the blocked signal 1003, and codes the vector pattern. Code sequence generation means with number as index signal 1015
Output to 1021. At this time, the input signal is an image showing the area shape information and the transparency information, and the intermediate level value is small, and most of the information is binary information. Therefore, when vector quantization is performed, the block containing the intermediate level value is included. Is. However, vector quantization may be performed when a vector in which all pixels have the maximum pixel value or all pixels have the minimum pixel value is prepared in the codebook and the intermediate level value is not included.

The code string generating means 1021 encodes the layered pattern signal 1020 and the index signal 1015 and outputs it as the encoded signal 1022. As shown in FIG. 21, when the pattern of the upper layer is known and 0 is included in the lower layer, 0 can be omitted, and by omitting 0, efficient compression can be realized.

As described above, according to this embodiment, the vector hierarchical coded three-dimensional pattern of the vector quantization codebook and the coded block hierarchical three-valued pattern are searched before searching the vector. By comparing and limiting the vector to be searched, high-speed vector quantization becomes possible.

FIG. 15 shows a block diagram of an image coding apparatus in an embodiment of the tenth invention. In the figure, 1001 is an input signal, 1002 is blocking means, 1003 is a blocking signal, 15
01 is a motion vector signal, 1502 is a motion vector output selection means, 1503 is a motion vector output signal, 1511 is a motion compensation means, 1512 is a motion compensation screen signal, 1513 is a motion compensation determination means, 1514 is an encoding control signal, and 1515 is Motion vector output control signal, 1516 additional information signal, 1521 transparency / area shape information coding means, 1522 coded signal, 1523 decoding means, 1524 decoding signal, 1525 memory means, 1526 reference It is a screen signal.

The operation of the image coding apparatus of the present embodiment configured as above will be described below.

In the block forming means 1002, the input signal 1001 which is the encoded signal is divided into blocks which are units of encoding processing, and the blocked signal 1003 is included in the transparency / area shape information encoding means 1521 and the motion compensation determining means. Output to 1511. As a coding unit, a rectangular block of 16 pixels in the vertical direction and 16 pixels in the horizontal direction and a rectangular block of 8 pixels in the vertical direction and 8 pixels in the horizontal direction are usually selected, but the encoding unit is not limited to this.

The motion compensating means 1511 moves / deforms the reference screen signal 1526 based on the motion vector signal 1501, and outputs the motion compensating screen signal 1512 to the motion compensation judging means 1513. At this time, the motion vector signal 1501 is often a motion vector representing a translation amount in block units, but may be a motion parameter representing rotation / enlargement / translation by affine transformation based on a linear model. Further, the motion vector signal 1501 may be obtained from the transparency / region shape information signal or may be obtained from the luminance signal and the color difference signal corresponding to the transparency / region shape information signal. Absent.

In the motion compensation judging means 1513, the motion compensation screen signal 1512 and the blocked signal 10 which is the coded signal.
If the difference value is less than or equal to a predetermined value, the motion vector output control signal 1515 is output to the motion vector output selecting means 1502, and at the same time, the motion compensation selection is output as the additional information signal 1516. If it is larger, the coding control signal 1514 is output to the transparency / region shape information coding unit 1521, and at the same time, the intra-screen coding selection is output as the additional information signal 1516.

At this time, if the predetermined value is increased, the probability that the motion compensation is selected increases, and only the motion vector and the additional information need be coded, so that the code amount can be suppressed. However, the error due to motion compensation increases accordingly. In addition, such switching of coding within / between screens has the following problems. In the screen in Fig. 27 /
The explanatory view regarding the problem of switching of encoding between screens is shown. As shown in the figure, even if the coded block has the minimum pixel value of all pixels, if the difference signal is less than or equal to a predetermined value, motion compensation is selected and a motion compensation error is accumulated. In this case, in particular, if all the pixels are the minimum, the coding amount is small in the coding of the transparency / region shape information. Therefore, the coding efficiency becomes better when the intra-screen coding is selected. Therefore, in the motion compensation determination means 1513, in the case of the minimum pixel value of all pixels and the maximum pixel value of all pixels, it is possible to improve encoding efficiency by adding a function of preferentially selecting intra-frame encoding. Moreover, it is possible to prevent the accumulation of errors due to motion compensation.

In the transparency / region shape information coding means 1521, when intra-picture coding is selected based on the coding control signal 1514, the blocking signal 1003 is described in inventions 1 to 9. It is encoded based on the encoding device described above and output as an encoded signal 1522.

In the decoding means 1523, in order to generate the motion compensation reference screen, the encoded signal 1522 is decoded, the decoded signal 1524 is output and stored in the memory means 1525, and at the time of encoding at the next time, , To the motion compensation means 1511 as a reference screen signal 1526.

The motion vector output selecting means 1502 outputs the motion vector output signal 1503 when the motion compensation is selected based on the motion vector output control signal 1515. Regarding the coding of the motion vector, it may be coded together with the luminance signal or the color difference signal corresponding to the transparency / region shape information which is the input signal, or may be coded by the motion vector output selecting means 1502. .

As described above, according to this embodiment, when the moving image of the transparency / region shape information is encoded, the determination is made based on the motion compensation difference signal, and when the difference is small, only the motion compensation is performed. If the difference is large, intra-frame coding is performed by the coding method of the above-mentioned image coding device, so that the time of the moving image can be calculated without preparing a vector quantization codebook for inter-screen coding. It is possible to perform efficient encoding using the correlation of directions.

[0160]

As described above, according to the present invention, the region shape information and the transparency information image, which are the input signals, do not contain much intermediate level values, and most of the pixels have the maximum pixel value or the minimum pixel value. By utilizing the fact that the value can be expressed by two values, the block containing the intermediate level value is vector-quantized, and the binary information that occupies most of the value is layered ternary pattern coding. Moreover, the practical effect is great in that efficient encoding and decoding can be realized.

In the image coding apparatus of the first invention,
Blocks containing intermediate level values are vector quantized,
By encoding the vector quantization pattern with a hierarchical ternary pattern, efficient encoding that matches the characteristics of the input image becomes possible.

In the image coding apparatus of the second invention,
By controlling the vector quantization of blocks containing intermediate level values and the hierarchical ternary pattern coding of binary information in parallel, it is possible to control the code amount of vector quantization based on the hierarchical structure of ternary patterns. Efficient by performing precise coding of the intermediate level value pattern by reducing the vector quantization processing block, or by performing variable length coding of the vector quantization code number and hierarchical ternary pattern. It becomes possible to perform good encoding of.

In the image decoding apparatus of the third invention,
By decoding the hierarchical ternary pattern coded signal, the binary information is decoded and the position of the vector quantized block is determined, and the vector quantized codebook is decoded according to the position. Since this is good, it is possible to perform decoding that does not require the coding position information of the vector quantization block as additional information.

In the image coding apparatus of the fourth invention,
The area shape information and the transparency information image, which are the input signals,
It is determined whether it is a binary image or a multi-valued image, vector quantization is performed only in the case of a multi-valued image, vector quantization is omitted in the case of a binary image, and only hierarchical ternary pattern coding is performed. By doing so, it is possible to realize a reduction in the amount of encoding processing and at the same time, to perform efficient encoding that matches the properties of the input image.

In the image coding apparatus of the fifth invention,
The area shape information and the transparency information image, which are the input signals,
Determine the type of image, whether it is a character image, a natural image, or a computer graphics image,
By preparing a vector quantization codebook for each type of image, efficient encoding that matches the nature of the image becomes possible.

In the image coding apparatus of the sixth invention,
Efficient encoding can be performed by analyzing the frequency of appearance of the pattern for each layer of the hierarchical ternary pattern and performing variable length encoding for each layer. Furthermore, by performing variable-length coding on the pattern of the lower layer according to the pattern of the upper layer, that is, by coding using the correlation between layers, efficient coding becomes possible.

In the image coding apparatus of the seventh invention,
By variable-length coding the vector quantization code number using the hierarchical ternary pattern, efficient coding can be performed by using the correlation between the hierarchical structure and the pattern of the vector quantization block. Furthermore, according to the hierarchical ternary pattern, the vector quantization block corresponding to the hierarchical pattern is rotated to concentrate the appearance frequency on a specific vector quantization pattern, and variable length coding is performed to achieve more efficient coding. Is possible.

In the image decoding apparatus of the eighth invention,
Performing decoding corresponding to the encoding device of the seventh invention, with respect to the rotation of the vector quantization block, since the rotation can be uniquely restored if the hierarchical ternary pattern is decoded,
It is possible to perform decoding without adding information about block rotation.

In the image coding apparatus of the ninth invention,
Before performing a vector search, the vector quantized codebook vector hierarchical ternary pattern is compared with the encoded block hierarchical ternary pattern to limit the vector to be searched. Vector quantization is possible.

In the image coding apparatus of the tenth invention,
When encoding a moving image of transparency / region shape information, determination is made based on a motion compensation difference signal. If the difference is small, only motion compensation is performed, and if the difference is large, the code of the image encoding device. By intra-coding by the coding method, it is possible to perform efficient coding using the correlation in the time direction of moving images without preparing a vector quantization codebook for inter-coding. .

[Brief description of drawings]

FIG. 1 is a block diagram of an image encoding device according to an embodiment of the first invention.

FIG. 2 is a block diagram of an image coding apparatus according to the first embodiment of the second invention.

FIG. 3 is a block diagram of an image coding apparatus according to a second embodiment of the second invention.

FIG. 4 is a block diagram of an image coding apparatus according to a third embodiment of the second invention.

FIG. 5 is a block diagram of an image coding apparatus according to a fourth embodiment of the second invention.

FIG. 6 is a block diagram of an image decoding apparatus according to an embodiment of the third invention.

FIG. 7 is a block diagram of an image coding apparatus according to an embodiment of the fourth invention.

FIG. 8 is a block diagram of an image coding apparatus according to an embodiment of the fifth invention.

FIG. 9 is a block diagram of an image encoding device according to a first embodiment of the sixth invention.

FIG. 10 is a block diagram of an image coding apparatus according to a second embodiment of the sixth invention.

FIG. 11 is a block diagram of an image coding apparatus according to a first embodiment of the seventh invention.

FIG. 12 is a block diagram of an image coding apparatus according to a second embodiment of the seventh invention.

FIG. 13 is a block diagram of an image decoding apparatus according to an embodiment of the eighth invention.

FIG. 14 is a block diagram of an image coding apparatus according to an embodiment of the ninth invention.

FIG. 15 is a block diagram of an image coding apparatus according to an embodiment of the tenth invention.

FIG. 16 is a block diagram of a conventional image encoding device by orthogonal transformation.

FIG. 17 is an explanatory diagram regarding an image hierarchical structure.

FIG. 18 is an explanatory diagram regarding an intermediate level value of transparency information.

FIG. 19 is an explanatory diagram regarding ternarization of a vector quantization pattern.

FIG. 20 is an explanatory diagram related to codeword generation of a ternary pattern.

FIG. 21 is an explanatory diagram regarding layering of a ternary pattern.

FIG. 22 (a) is an explanatory diagram of a pattern in which a ternary pattern frequently appears, and FIG. 22 (b) is an explanatory diagram of a pattern in which a ternary pattern rarely appears.

FIG. 23 (a) is an explanatory diagram regarding the appearance frequency of a ternary pattern by correlation with surrounding blocks, and (b) is an explanatory diagram regarding the appearance frequency of vector quantization code patterns by correlation with surrounding blocks.

FIG. 24 is an explanatory diagram of variable-length coding using directional correlation between layers.

FIG. 25 is an explanatory diagram regarding a method of determining a directional state in an upper layer.

FIG. 26 is an explanatory diagram regarding a method for accelerating a vector search using a hierarchical ternary pattern.

[Fig. 27] Fig. 27 is an explanatory diagram regarding a problem of intra-screen / inter-screen encoding switching

[Explanation of symbols]

 1002 Blocking Means 1004 Vector Quantizing Means 1007 Three-Valued Means 1009 Hierarchical Means 1011 Code Sequence Generation Means 1014 Vector Quantization Means 1017 Three-Valued Means 1019 Hierarchical Means 1021 Code Sequence Generation Means 1023 Code Amount Calculation Means 1031 Small Blocks Coding means 1041 hierarchized pattern variable length coding means 1043 code string generation means 1044 index variable length coding means 1051 binary / multivalued judgment means 1053 selection means 1054 selection means 1061 image judgment means 1065 codebook selection means 1071 hierarchization means 1073 Layer 1 pattern variable length coding means 1076 Layer 2 pattern variable length coding means 1078 Code string generation means 1081 Index variable length coding means 1083 Layered pattern variable length coding means 1086 Code string generation means 1104 Vector quantization means 1107 Ternarization means 1109 layering means 1111 code string generation means 1204 vector quantization means 1211 code string generation means 1301 block rotation Means 1404 Codebook ternarization means 1406 Codebook hierarchy means 1408 Hierarchical pattern comparison means 1502 Motion vector output selection means 1511 Motion compensation means 1513 Motion compensation determination means 1521 Transparency / region shape information coding means 1523 Decoding means 1525 Memory means 2002 Code string decoding means 2004 Vector dequantization means 2007 Three-value pattern generation means 2009 Decoded signal generation means 2102 Code string decoding means 2104 Vector dequantization means 2107 Three-valued pattern generation means 2109 Decoded signal generation means 2111 Rotation state determination means 2113 Block rotation means

Claims (18)

[Claims] 1. Blocking means for dividing an input signal, which is an image to be coded, into blocks, which is a signal representing image transparency information, area shape information, etc., and outputs as a blocked signal. The block signal is vector quantized in block units, and a vector quantization pattern signal representing the state of vector quantization and an index signal representing the index of the selected vector are respectively supplied to the ternarization means and the code string generation means. Vector quantizing means to output, the vector quantization pattern signal is divided into three states, a block to which an index is assigned, a block in which all pixels have a minimum pixel value, and a block in which all pixels have a maximum pixel value. , Ternary, output as ternary pattern signal
The ternarization means, and the ternary pattern signal, a plurality of blocks in which all pixels are the minimum pixel value and a plurality of blocks in which all the pixels are maximum pixel value are respectively layered together, and output as a layered pattern signal. The layering means for, the layering pattern signal and the index signal,
An image coding apparatus comprising: a code string generating unit that codes based on a predetermined condition to generate a code string and output the code string as a coded signal. 2. A signal representing image transparency information, region shape information, etc., which is an image to be coded, is divided into blocks for each processing unit, and as a blocked signal, a ternarization means and vector quantization are performed. Blocking means for outputting to the means, vector quantization of the blocked signal in block units, and vector quantizing means for outputting an index signal representing the index of the selected vector to the code string generating means, the blocking means A signal is divided into three states, that is, a block in which all pixels have a minimum pixel value, a block in which all pixels have a maximum pixel value, and another block. The three-valued pattern signal is hierarchically grouped into a plurality of blocks in which all pixels have a minimum pixel value and a plurality of blocks in which all pixels have a maximum pixel value, And layering means for outputting as a layering pattern signal, the hierarchical pattern signal and said index signal,
An image coding apparatus comprising: a code string generating unit that codes based on a predetermined condition to generate a code string and output the code string as a coded signal. 3. A vector quantization control signal which receives the hierarchical pattern signal as input, calculates the code amount of the hierarchical pattern signal, and reduces the code of a vector index when the code amount exceeds a predetermined code amount. 3. The image coding apparatus according to claim 2, further comprising code amount calculation means for outputting to the quantization means. 4. Before the vector quantizing means, the blocking signal output from the blocking means is used as an input signal, divided into blocks smaller than the blocking means, and the vector quantizing means is supplied to the vector quantizing means. 3. The image encoding device according to claim 2, further comprising a small block forming unit that outputs a small block forming signal. 5. The hierarchical pattern signal as an input signal,
Variable length coding is performed using a predetermined coding method, a hierarchical pattern variable length coding means for outputting a hierarchical pattern variable length coded signal to the code string generating means, and the index signal as an input signal, 5. Variable length coding is performed using a predetermined coding method, and index variable length coding means for outputting an index variable length coding signal to the code string generating means is added. One of the image encoding devices. 6. A signal encoded by the image encoding device according to claim 1 or 2 is used as an input signal, and is separated into a hierarchical pattern decoding signal and an index decoding signal, each of which is a ternary pattern. A code string decoding means for outputting to the generation means and the vector dequantization means, and reproducing a hierarchical structure from the hierarchical pattern decoded signal,
A ternary pattern generating means that develops a ternary pattern decoded signal by expanding it into a value pattern, searches a vector from the index decoded signal and allocates it, generates a decoding screen for each processing block, and performs block decoding Vector dequantization means for outputting a quantized signal, the ternary pattern decoded signal and the block decoded signal as an input signal, from the ternary pattern decoded signal, to determine the position of the vector quantized block , The block decoding signal is assigned to that position, and in the case of blocks other than the vector quantization block, the minimum pixel value or the maximum pixel value is assigned to all the pixels of the block according to the value of the ternary pattern decoding signal. And a decoded signal generating means for outputting the decoded signal as a decoded signal. 7. Blocking means for dividing an input signal, which is an image to be coded, into blocks, which is a signal representing image transparency information, area shape information, etc., and outputs as a blocked signal. A binary / multivalue determination means for determining whether the blocking signal is a binary signal or a multivalued signal and outputting a selection control signal to the first selecting means and the second selecting means; Based on the selection control signal, the first signal is output to the second selection means when the selection control is binary, and to the vector quantization means when the selection control is multivalued. Selecting means, and the blocking signal output from the first selecting means,
Vector quantization in which vector quantization is performed in block units, and a vector quantization pattern signal representing the state of vector quantization and an index signal representing the index of the selected vector are respectively output to the second selecting means and the code string generating means. Based on the selection control signal, if the selection control is binary output the blocking signal as a selection blocking signal,
When the selection control is multi-valued, second selection means for outputting the index signal as a selection block signal, and the selection block signal, a block in which all pixels are minimum pixel value, all pixels are maximum pixel value One block, divided into three states of other blocks, ternarization means for ternarizing and outputting as a ternary pattern signal, the ternary pattern signal, a plurality of blocks in which all pixels are the minimum pixel value And a plurality of blocks in which all the pixels have the maximum pixel value are hierarchized collectively, and a hierarchizing means for outputting as a hierarchized pattern signal, the hierarchized pattern signal and the index signal,
An image coding apparatus comprising: a code string generating unit that codes based on a predetermined condition to generate a code string and output the code string as a coded signal. 8. Blocking means for dividing an input signal, which is an image to be coded, into blocks, which is a signal representing image transparency information, area shape information, and the like, and outputs as a blocked signal. An image determination unit that classifies the blocked signal based on a predetermined condition and outputs a codebook selection control signal to the codebook selection unit; and a selection from at least two codebooks based on the codebook selection control signal. Codebook selecting means for outputting a codebook, vector-quantizing the blocked signal in block units with reference to the selected codebook, and a vector quantization pattern signal representing the state of vector quantization, and selected. An index signal representing the index of the vector is output to the ternarization means and the code string generation means, respectively. Vector quantization means, and the vector quantization pattern signal, divided into three states of a block to which an index is assigned, a block in which all pixels have a minimum pixel value, a block in which all pixels have a maximum pixel value, 3 Quantize and output as a ternary pattern signal
The ternarization means and the ternary pattern signal are hierarchically grouped into a plurality of blocks in which all pixels have the minimum pixel value and a plurality of blocks in which all the pixels have the maximum pixel value, and output as a hierarchical pattern signal. The layering means for, the layering pattern signal and the index signal,
An image coding apparatus comprising: a code string generating unit that codes based on a predetermined condition to generate a code string and output the code string as a coded signal. 9. The image determining means measures a frequency distribution of pixel values, and the codebook selecting means is suitable for a binary image when the intermediate level value is small in accordance with the ratio of the intermediate level value distribution. 9. The image coding apparatus according to claim 8, wherein control is performed to select a codebook that has been converted into a codebook and a codebook that is optimized for a multi-valued image when there are many intermediate level values. 10. A signal representing image transparency information, area shape information, etc., which is an input signal to be coded, is divided into blocks, and a ternarization means and vector quantization are performed as a blocked signal. Blocking means for outputting to the means, vector quantization of the blocked signal in block units, and vector quantizing means for outputting an index signal representing the index of the selected vector to the code string generating means, the blocking means A signal is divided into three states, that is, a block in which all pixels have a minimum pixel value, a block in which all pixels have a maximum pixel value, and another block. The three-valued pattern signal is hierarchically grouped into a plurality of blocks in which all pixels have a minimum pixel value and a plurality of blocks in which all pixels have a maximum pixel value. Layering means for outputting a layer 1 pattern signal and a layer 2 pattern signal for each layer, and layer 1 pattern variable length encoding means for variable length encoding the layer 1 pattern signal and outputting it as a layer 1 pattern variable length encoded signal. A layer 2 pattern variable length coding means for variable length coding the layer 2 pattern signal and outputting as a layer 2 pattern variable length coded signal; the layered 1 pattern variable length coded signal;
An image coding apparatus comprising a two-pattern variable-length coded signal and a code string generation means for coding the index signal based on a predetermined condition to generate a code string and output the code string as a coded signal. 11. The layer 1 pattern variable length coding means performs variable length coding on the layer 1 pattern signal, outputs the layer 1 pattern variable length coded signal, and analyzes the layer 1 pattern signal under a predetermined condition. Then, the function of outputting the coding control signal to the layer 2 pattern variable length coding means is added, and in the layer 2 pattern variable length coding means, the layer 2 pattern signal is based on the coding control signal, Analyze the correlation between layer 1 and layer 2 and perform variable length coding
A function of outputting as a pattern variable length coded signal is added, and in the code string generation means, the layer 1 pattern variable length coded signal, the layered 2 pattern variable length coded signal, and the index signal are included. 11. The image coding apparatus according to claim 10, wherein the image coding apparatus has a function of performing coding based on a predetermined condition, generating a code string, and outputting the code string as a coded signal. 12. A signal representing image transparency information, area shape information, etc., which is an input signal to be coded, is divided into blocks, and a ternarization means and vector quantization are performed as a blocked signal. A block quantizing means for outputting to the means, a vector quantizing means for vector-quantizing the blocked signal in block units, and outputting an index signal representing an index of the selected vector, the blocked signal for all pixels A block having the minimum pixel value, a block having all the pixels having the maximum pixel value, and a case in which the state is divided into three states, and a ternarizing means for ternarizing and outputting as a ternary pattern signal, the ternary pattern The signal is hierarchically grouped into a plurality of blocks in which all pixels have a minimum pixel value and a plurality of blocks in which all pixels have a maximum pixel value, and each layer is layered. A layering means for outputting a turn signal, variable length coding the layered pattern signal and outputting it as a layered pattern variable length coded signal, analyzing the layered pattern signal under a predetermined condition, and index variable length A hierarchical pattern variable length coding means for outputting a coding control signal to a coding means; and a variable length code for analyzing the correlation between the hierarchical pattern and the vector index based on the index control signal for the index signal. Vector index variable-length coding means for generating a variable-length coded signal and outputting the layered pattern variable-length coded signal and the index signal based on a predetermined condition to generate a code string. And an image coding device configured to output a coded signal as a coded signal. 13. The layered pattern variable length coding means performs variable length coding on the layered pattern signal, outputs the layered pattern variable length coded signal, and analyzes the layered pattern signal under a predetermined condition. Then, the rotation control signal is output to the block rotation means, while the function of outputting the coding control signal to the index variable length coding means is added, and the blocking signal is based on the rotation control signal. 13. A block rotating means for rotating a block at a predetermined angle and outputting it as a rotation block signal is added.
The image encoding device described. 14. The hierarchical pattern variable length coding means outputs the hierarchical pattern signal to the block rotation means as a control signal such that the distribution of maximum pixel values is biased to the lower right. Image coding device. 15. A signal encoded by the image encoding device according to claim 12 is used as an input signal, and is separated into a hierarchical pattern decoded signal and an index decoded signal, and the hierarchical pattern decoded signal is a ternary pattern. A code sequence decoding means for outputting the index decoded signal to a vector dequantization means in the generation means and the rotation state determination means, and reproducing a hierarchical structure from the hierarchical pattern decoded signal, 3
A ternary pattern generating means that develops a ternary pattern decoded signal by expanding it into a value pattern, searches a vector from the index decoded signal and allocates it, generates a decoding screen for each processing block, and performs block decoding Vector dequantization means for outputting a coded signal, analyzing a pattern from the hierarchical pattern decoded signal,
A rotation state determination unit that analyzes the rotation state of the vector quantization block and outputs a rotation control signal, and the block decoded signal that rotates the block based on the rotation control signal and outputs the block rotation decoded signal. Block rotation means, using the ternary pattern decoded signal and the block rotation decoded signal as an input signal, from the ternary pattern decoded signal, determine the position of the vector quantized block, the block at the position In the case of a block other than the vector quantization block, the rotation decoding signal is assigned, and the decoding signal is obtained by substituting the minimum pixel value or the maximum pixel value for all the pixels of the block according to the value of the ternary pattern decoding signal And an image decoding device including a decoded signal generating means for outputting as. 16. A signal representing image transparency information, area shape information, etc., which is an input image to be coded, is divided into blocks, and a ternarization means and vector quantization are performed as a blocked signal. Blocking means for outputting to the means, the blocking signal, the case where all pixels are blocks having the minimum pixel value, all pixels are blocks having the maximum pixel value, and cases are divided into three states, and three-valued And a ternarization unit that outputs the ternary pattern signal as a ternary pattern signal, the ternary pattern signal, a plurality of blocks in which all pixels are minimum pixel values and a plurality of blocks in which all pixels are maximum pixel values And a codebook output by the vector quantizing means, and a hierarchizing means that outputs the hierarchized pattern signal to the code string generation means and the hierarchized pattern comparison means.
A codebook ternarization means for categorizing into three states, that is, a block in which all pixels have the minimum pixel value, a block in which all pixels have the maximum pixel value, and another block, and outputs as a codebook ternary pattern signal. And, the codebook ternary pattern signal is hierarchically grouped into a plurality of blocks in which all pixels have a minimum pixel value and a plurality of blocks in which all pixels have a maximum pixel value. The codebook layering means for outputting to the layered pattern comparison means, the layered pattern signal and the codebook layered pattern signal are compared, the correlation of the layered pattern is analyzed, and the vector quantization control signal is output. And a layered pattern comparison means for performing vector quantization on the block basis on the basis of the vector quantization control signal, and selecting the block signal. And an index signal representing the index of the vector and outputs the code string generation means, and a vector quantization means for outputting a code book in the code book ternary means, the layering pattern signal and said index signal,
An image coding apparatus comprising: a code string generating unit that codes based on a predetermined condition to generate a code string and output the code string as a coded signal. 17. A signal representing transparency information and area shape information of an image, wherein an input signal which is an image to be coded is divided into blocks, and the transparency / area shape information coding is performed as a blocked signal. Means and a motion compensation judging means, a block compensating means for outputting to the means and the motion compensation judging means, a motion compensating means for moving and transforming the reference screen signal based on the motion vector signal, and outputting the motion compensating screen signal, If the difference value is less than or equal to a predetermined value, the motion vector output control signal is output to the motion vector output selecting means, and at the same time, the motion compensation selection is output as an additional information signal, and the difference value is larger than the predetermined value. For example, a motion compensation determination means for outputting an encoding control signal to the transparency / area shape information encoding means and at the same time outputting an intra-frame encoding selection as an additional information signal, When intra-picture coding is selected based on the coding control signal, the blocked signal is coded based on the coding device described in claims 1 to 16 and output as a coded signal. Transparency / region shape information coding means, decoding means for decoding the coded signal and outputting the decoded signal to the memory means, storing the decoded signal, and coding at the next time, An image composed of memory means for outputting to the motion compensating means as a reference screen signal, and motion vector output selecting means for outputting a motion vector output signal when motion compensation is selected based on the motion vector output control signal. Encoding device. 18. The motion compensation selection means does not select motion compensation when all pixel values of the blocked signal which is a block to be coded are the maximum value or the minimum value, and the intra-frame code is not selected. 18. The image encoding device according to claim 17, further comprising a function for selecting encoding.