JP2000333016A - Image coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To eliminate a background noise caused by a near lossless coding without provision of a specific configuration. SOLUTION: Alocal decoder 114 in a near lossless coder 100 decodes a quantized signal and gives the result to a clamp circuit 160 in order that a dynamic range of a decoded image is adjusted not excess of a dynamic range of an original image. A threshold (clamp start point) of this clamp circuit 160 is set to a value other than a maximum value(MAXVAL) and a minimum value (0) of an original image, and dispersion in a pixel value caused by a quantization error is absorbed. Concretely, in the case that a parameter to decide a permissible maximum width of a quantizer 130 is set to NEAR, the clamping threshold is set to 'MAXVAL-NEAR' and 'NEAR'.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding apparatus for efficiently compressing / expanding a halftone image, and more particularly, to a near lossless code capable of increasing a compression ratio while suppressing visual deterioration of a restored image to a small extent. Device.

[0002]

2. Description of the Related Art A "halftone image" having a high adjacent correlation such as a photograph
The JPEG method is suitable for encoding, but JPEG is not suitable for a “character image” including many sharp edges because the image quality deteriorates conspicuously. On the other hand, when the binarization (JBIG) method is used in consideration of only the character image, halftone information of the halftone image is lost, and the image quality is deteriorated.

[0003] For this reason, in an apparatus that demands high restoration image quality for "character images and mixed images (images in which character images and halftone images are mixed)", lossless coding and visual deterioration are reduced. Suppressed near lossless coding is suitable.

[0004] Near-lossless coding means that although an error is unavoidable due to the quantization involved in the coding process (lossy coding), the error is reduced so as to hardly affect human vision. For example, compression coding according to the ISO (International Organization for Standardization) JPEG-LS standard is applicable.

[0005] Near-lossless encoding of this ISO standard is based on J
Unlike PEG, it is intended to store and compress the pixel value information of an image as it is, without performing DCT transformation, and basically to quantize the prediction error and then perform entropy coding. is there. In the case of near-lossless coding, there is an advantage that halftone information can be stored as it is.

For example, in a FAX multifunction machine that emphasizes a copy function, a halftone image read by a scanner is near-losslessly compressed and temporarily stored in a memory, and when printing, the compressed data is restored so as to be suitable for a printer. Necessary image processing such as resolution conversion and error diffusion processing is performed and output. According to such a method, there is an advantage that it is easy to obtain a processed image of high image quality because halftone information is stored, as compared with processing a binarized image.

FIG. 6 is a block diagram showing a main configuration of a near-lossless encoder (prediction encoder) of the ISO standard. Hereinafter, the characteristic operation of this device will be briefly described.

[0008] The image data is read into the line memory 210. The line memory 210 includes a coding line and a reference line,
It is composed of two line memories. Predictor 220
Predicts the value of a coded pixel from the state of neighboring pixels of the coded pixel.

FIG. 7 shows an example of the arrangement of encoded pixels and peripheral pixels. In the figure, “Ix” represents the level (pixel value) of the coded pixel, and here, it takes a value in the range of “0 to 255”. “Ra”, “Rb”, “Rc”, and “Rd” similarly represent the levels of four reference pixels.

[0010] The four reference pixels are input to the predictor 220, and the value of "Ix" is predicted based on the four reference pixels. Forecasted value is "P
x ”.

In the subtractor 221, the prediction error signal Errval = Ix
Calculate Px. In the case of lossless coding, Errval is encoded without quantization, but in the case of near lossless coding, this value is quantized so that the compression ratio can be increased.
Quantize at 230 and encode.

The point to be noted here is that, when quantizing,
The maximum width of the allowable error range is determined by the control parameter (NEA
R). That is, "NEAR" is a parameter indicating the degree of quantization, and the value after quantization is guaranteed to be in the "range of ± NEAR based on the value before quantization".

In the C language notation, quantization is expressed as follows. If (Errval> 0) Errval = (Errval + NEAR) / (2 * NEAR + 1) else Errval = − (NEAR Errval) / (2 * NEAR +1) Since Errval is an integer, for example, Errval = 15, NEAR = Three
If so, Errval after quantization is truncated to "2"
Becomes This value is encoded by the entropy encoder 240.

On the other hand, at the time of predictive encoding, it is necessary to perform encoding using a reference pixel having the same value as a reference pixel used at the time of decoding in order to achieve matching with decoding. Therefore, as shown in FIG.
202, the output of the quantizer 230 is decoded, and then the minimum value “0” and the maximum value “MAXVAL” of the pixel value of the original image are set so as not to exceed the dynamic range of the original image. As a value (clamp start point), the output of the local decoder 202 is clamped to the minimum value “0” and the maximum value “MAXVAL” of the pixel value of the original image to restore the reference pixel.

That is, assuming that the output of the quantizer 230 is Errval, when this is inversely quantized by the inverse quantizer 250,
rrval = Errval * (2 * NEAR + 1). In the above numerical example, Errval output from the decoder is the encoded value "2"
Therefore, after inverse quantization, Errval becomes “14”.

The predicted value "Px" is added to the inversely quantized value for decoding. In other words, the value after addition is Rx = Px
+ Errval * (2 * NEAR + 1). This addition is performed by the adder 251 of the circuit of FIG.

The “Rx” obtained in this manner may vary in the range of ± NEAR as compared with the original pixel “Ix”. Therefore, if an error of ± NEAR occurs near the maximum value or the minimum value of the pixel value, “Rx” may exceed the maximum value of the original image or fall below the minimum value. That is, the case where the dynamic range of the pixel value of the restored image exceeds the dynamic range of the original image is also assumed.

In order to prevent this, a clamp circuit 260 is provided. If the dynamic range of the pixel value of the original image is “0 to MAXVAL”, the clamp circuit 260
Clamps the restored pixel signal “Rx” with “0” and “MAXVAL” as thresholds. That is, when "Rx" takes a value equal to or less than "0", it is clamped to "0", and when it takes a value equal to or more than "MAXVAL", it is clamped to "MAXVAL". That is, in the C language notation: In this embodiment, MAXVAL = 255. “Rx” output from the clamp circuit 260 is sequentially as shown in FIG.
The data is written to the line memory 210, and this becomes a reference pixel for encoding the next pixel.

[0019]

In the near-lossless coding technique described above, the pixels of level "0" of the original image are "0-NE".
AR "range. Pixels having MAXVAL are restored to the range of “MAXVAL-NEAR to MAXVAL”.

For example, when one pixel is represented by 8 bits, a black pixel normally has a level “0” and a white pixel has a level “255”. Assuming that NEAR is "3", black pixels are "0 to 3", white pixels are "252 to 255", and the background color is slightly gray.

If this restored signal is subjected to error diffusion processing (halftone processing) so as to be output to a binary printer, a problem arises particularly when the background color is white. That is, a black dot corresponding to the level of “252” appears in the background (white background), and the background becomes a background in which black dots are mixed. It should be noted that when the foreground is black, there is often little visual problem that the black becomes slightly lighter.

The above problem can be solved by performing a gamma correction process or the like. In this case, however, a special correction circuit is required, and new problems such as an increase in processing time and an increase in circuit scale are required. Occurs.

The present invention has been made in view of such a problem, and it is an object of the present invention to eliminate background noise caused by a quantization error in near lossless encoding without using a special configuration. Aim.

[0024]

According to the present invention, the upper limit value or lower limit value of the original pixel is reduced by a quantization error due to a very simple method of changing the threshold value (clamp start point) of the clamp circuit. Prevents dots from appearing in the vicinity.

For example, if the threshold values of the clamp are "0" and "M
AXVAL "to" NEAR "," MAXVAL-NEAR "
All levels from "0 to NEAR" are set to "0" and "MAXVAL-NEA"
By setting all the levels of “R to NEAR” to “NEAR”, variations in pixel values caused by quantization errors are absorbed. Therefore, even when image processing such as error diffusion is performed, a situation where black dots are mixed in the background color (white) does not occur.

[0026]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In a first embodiment of the image coding apparatus according to the present invention, a picture is generated based on a pixel value of a pixel to be coded and a pixel previously coded and restored by a local decoder. Encoding means for performing an entropy encoding after quantizing an error with a predicted value, and a clamp functioning to adjust a dynamic range of an output of the local decoder so as not to exceed a dynamic range of pixel values of an original image. Means for adjusting the clamping start point in the clamping means so that the background or foreground value decoded by the local decoder is the same as the background or foreground value of the original image. Control.

By adjusting the starting point of the clamp, the upper limit of the value of the original pixel. Form a dead zone near the lower limit,
Absorbs pixel value variations. That is, even when the signal value of the restored pixel is slightly smaller than the maximum pixel value, the variation is eliminated by forcibly unifying the signal value to the maximum pixel value. As a result, the background value (maximum pixel value) and the foreground value (minimum pixel value) of the restored image match the original image. Therefore, even if error diffusion processing or the like is performed after decoding, the background can be white.

According to a second aspect of the present invention, in a predictive coding apparatus for predicting and coding the value of a pixel to be coded from the value of a reference pixel, a quantizer for quantizing a prediction error; An entropy encoder for encoding the output of the quantizer, a local decoder for decoding from the prediction error and the prediction value quantized by the quantizer, and an output value of the local decoder for a predetermined value. When the output value reaches the threshold value, the output value is clamped to the maximum value or the minimum value of the pixel value of the original pixel. The pixel value is a value other than the maximum value or the minimum value.

According to such a configuration, variations in pixel values are absorbed by using the clamping process in local decoding, so that background noise can be removed without using a special circuit.

In a third aspect of the present invention, the quantizer can control the level of the quantization error by the value of a control parameter. As a result, so-called near-lossless encoding can be realized while suppressing an error due to quantization within a given range.

In the fourth aspect of the present invention, the control parameter is a value defining a maximum width of an allowable quantization error, and the predetermined value in the clamping means is set to a value of an original image. A value obtained by subtracting the value of the maximum width of the allowable quantization error from the maximum value of the pixel value, or a value obtained by adding the value of the maximum width of the allowable quantization error to the minimum value of the pixel value of the original image. I do.

Thus, near the upper limit and lower limit of the pixel value of the original image (within the range of the allowable quantization error), the pixel value of the restored image becomes the upper limit and lower limit of the pixel value of the original image. Thus, the foreground value and the background value of the original image and the restored image match, and the image quality does not deteriorate.

In a fifth aspect of the present invention, the present invention is applied to a near-lossless image coding apparatus, which can store halftone information as it is, has a high compression ratio, and prevents image quality deterioration of a restored image. A possible encoding is realized.

Also, the image decoding apparatus according to the sixth aspect of the present invention decodes by performing a process reverse to the encoding.
A decoded signal is obtained by clamping the decoded signal under the same conditions as the clamping conditions used for encoding.

By employing the same clamp conditions as in the case of encoding, decoding can be performed using the same reference pixels as those used for encoding.

A communication device according to a seventh aspect of the present invention performs an image encoding device, an image decoding device, a memory for storing encoded or decoded signals, and performs predetermined image processing. And an image processing device.

According to this communication device, the halftone image is compressed with high efficiency without losing the information and stored in the memory, and at the time of printing or the like, the code data stored in the memory is decoded and Processing such as executing resolution conversion or error diffusion processing so as to be suitable for a printer to obtain a high-quality restored image is possible.

In the image coding method according to the eighth aspect of the present invention, the image is generated based on the pixel value of the pixel to be coded and the pixel previously coded and restored by the local decoder. An image coding method for performing entropy coding after quantizing an error from a prediction value, wherein the local decoding is performed such that a dynamic range of an output signal of the local decoder does not exceed a dynamic range of pixel values of an original image. When clamping the output signal of the device, the clamp start point is set to a value other than the maximum or minimum pixel value of the original image.

According to this method, the background noise can be removed only by changing the clamp start point (clamp threshold).

Further, in the image coding method according to the ninth aspect of the present invention, the maximum width of an allowable quantization error is specified by a control parameter at the time of the quantization.
The clamp start point is defined as a value obtained by subtracting the maximum value of the allowable quantization error from the maximum value of the pixel value of the original image, or the minimum value of the pixel value of the original image. The value of the maximum width is added.

A method in which the threshold value of the clamp is changed to form a dead zone near the upper limit value and the lower limit value of the pixel value of the original image (within an allowable quantization error range) to absorb the variation in the pixel value. It is.

Hereinafter, embodiments of the present invention will be described more specifically with reference to the drawings.

FIG. 1 is a block diagram showing a main configuration of a communication apparatus (such as a facsimile apparatus) to which the present invention is applied.

As shown in the figure, the communication apparatus comprises a predictive encoder 100 and a memory 170 for temporarily storing code data.
, A decoder 180, an image processing circuit 190 for performing resolution conversion, error diffusion processing, and the like, and a printer 192.

The basic structure of the near lossless encoder (prediction encoder) 100 is the same as that of the conventional example shown in FIG.

That is, a local decoder (including an inverse quantizer 150 and an adder 151) 114, a line memory 116,
A predictor 120, a subtractor 121, a quantizer 130 capable of defining a maximum width of an error, an entropy encoder (lossless encoder) 140 such as an arithmetic code or a Golomb code, and a clamp circuit 160
And

It should be noted that the clamp circuit 16
The threshold value at 0 (the value at which clamping starts) is shown in FIG.
"NEAR" and "MAXVAL-NE" instead of "0" and "MAXVAL"
AR ”.

As described above, "NEAR" is
A control parameter that defines the degree of quantization in the quantizer 130, and by appropriately setting NEAR,
The value after quantization is guaranteed to fall within the range of ± NEAR based on the value before quantization.

When the threshold value of the clamp circuit 160 is changed as shown in FIG. 1, the input and output of the clamp circuit 160 become
The relationship shown in FIG. 3C is established, and a flat characteristic (dead zone) in which the output does not change with respect to the input is formed within the dynamic range of the pixel value of the original image.

That is, the vicinity of the maximum pixel value “255” (255 to
NEAR), a flat characteristic is formed, and a flat characteristic is formed around the minimum pixel value “0” (0 to NEAR). With this flat characteristic, it is possible to remove an unnecessary pixel distribution that occurs in a restored image due to a quantization error.

That is, when an image is decoded by the local decoder, as shown in FIG.
A pixel distribution that does not exist originally (indicated by an arrow in the drawing) appears near the minimum value. If error diffusion processing or the like is performed in this state, the image quality is degraded by black being mixed with white on the background. Therefore, by changing the threshold value of the clamp circuit, a characteristic as shown in FIG. , All pixels having a value of “255 to NEAR” are integrated into a pixel of “255”, and “0 to
All pixels having a value of "NEAR" are integrated into a pixel of "0". As a result, the distribution of the pixel values of the restored image is shown in FIG.
As shown in (a), the image coincides with the original image. Therefore, even if image processing is performed thereafter, the image does not deteriorate.

Hereinafter, the operation of the prediction encoder 100 shown in FIG. 1 will be specifically described.

First, image data to be encoded is read into the line memory 116. The line memory 116 includes an encoding line, a reference line, and two line memories, as in the conventional example (FIG. 7).

The four reference pixels are input to the predictor 120, and the value of Ix (FIG. 7) is predicted based on the four reference pixels. The predicted value is set to Px.

In the subtractor 121, the prediction error signal Errval = Ix
Calculate Px. Next, the value is quantized by the quantizer 130. As described above, the maximum allowable error can be defined by the value of the parameter “NEAR” representing the degree of quantization.

The quantization processing is expressed in C language as follows. If (Errval> 0) Errval = (Errval + NEAR) / (2 * NEAR + 1) else Errval = − (NEAR Errval) / (2 * NEAR +1) And the output value of quantizer 130 is the entropy encoder 140
. The entropy encoder 140, for example,
It can be realized by an arithmetic coder or a Golomb coder.

The local decoder 114 generates a pixel value Rx which is restored when a decoding process is performed later. Therefore,
Rx = Px + Errval * (2 * NEAR + 1).

This “Rx” is input to the clamp circuit 160, where a clamp process is performed. The operation of the clamp circuit 160 is described below in C language. By performing such a clamping process, pixels from level “0 to NEAR” are clamped to level “0”,
Pixels with values from "MAXVALNEAR to MAXVAL"
AL ”. Therefore, this processing can remove the pale background color generated in the conventional example. Rx after the clamping process is sequentially written to the line memory 116,
This becomes a reference pixel for encoding the next pixel.

In the communication device shown in FIG. 1, encoded data is temporarily stored in memory 170. Then, at the time of printing, the stored data is read out, and the decoder 180 performs a decoding process.

FIG. 2 shows a configuration example of the decoder 180. As shown, the decoder 180 performs processing in the reverse order of the encoder 100, and the entropy decoder 181
, An inverse quantizer 182, a predictor 183, an adder 184, a clamp circuit 185, and a line memory 186.

Hereinafter, the decoding operation will be described.

Since there is no loss of information between the entropy encoder 140 and the entropy decoder 181, an encoded Errval is obtained from the entropy decoder 181. When it is inversely quantized by the inverse quantizer 182, as in the conventional example, E
rrval = Errval * (2 * NEAR + 1) is obtained.

The predictor 183 has the same configuration as the predictor 120 inside the encoder. Assuming that the predicted value is Px, the value after addition is
Rx = Px + Errval * (2 * NEAR + 1). The clamp circuit 185 performs the following processing as in the case of the local decoder in the encoder. By doing so, the pixels from level “0 to NEAR” are clamped to level “0” and “MAXVALNEAR to MAXVA”
Pixels with values up to "L" are clamped to "MAXVAL". Therefore, this processing can remove the pale background color generated in the conventional example. "Rx" sequentially, line memory
Writing to 186 completes the decoding of one line.

FIG. 4 shows an example of the hardware configuration of the facsimile apparatus having the configuration shown in FIG.

As shown, the facsimile apparatus 101
Is a host processor 10 that controls the operation of each part
2, a near-lossless encoding / decoding circuit 103, an image processing circuit 104 for performing resolution conversion and error diffusion processing, a QM encoding / decoding circuit (arithmetic encoding / decoding circuit) 105, and an image line memory 106. , A code memory 107, a communication interface circuit 108 such as a modem for transmitting and receiving signals to and from a destination communication device via a communication line 113, an image input device 111 such as a scanner, and an image recording / display device such as a printer. 112
And Reference numerals 109 and 110 are internal buses.

In this embodiment, the near lossless encoding /
It is only necessary to change the threshold value of the clamp circuit in the decoding circuit 103, and no special circuit such as a gamma correction circuit is required. Therefore, the configuration of the device is not complicated, the speed of signal processing is not reduced, and the implementation is easy.

Although FIG. 1 shows a configuration necessary for storing the encoded data in the memory and restoring it at the time of printing, the encoded data is encoded by the encoder 100 as shown in FIG. The data can be transmitted to the communication device of the other party via the communication line L1, and can be decoded by the decoder 180 in the other device. In this case, the clamp condition of the clamp circuit 185 on the reception side is the same as the clamp condition of the clamp circuit 160 on the transmission side.

[0068]

As described above, according to the present invention, it is possible to eliminate background noise caused by near lossless coding without adding a special circuit. Therefore, it is possible to freely perform resolution conversion, error diffusion processing, and the like on the near-lossless encoded data without worrying about image quality deterioration.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a main configuration of a communication device according to an embodiment of the present invention;

FIG. 2 is a block diagram showing a configuration example of a decoder according to the embodiment of the present invention;

FIG. 3A is a diagram illustrating an example of a distribution of pixel values of an image to be encoded; FIG. 3B is a diagram illustrating an example of a distribution of pixel values of a decoded image; FIG. For explaining that a dead zone is formed at a desired level of the output of the clamp circuit.

FIG. 4 is a block diagram showing a hardware configuration example of a facsimile apparatus to which the present invention is applied;

FIG. 5 is a block diagram illustrating a configuration example of a communication system in which an encoding circuit is provided on a transmission side and a decoding circuit is provided on a reception side.

FIG. 6 is a block diagram showing a configuration of a conventional example.

FIG. 7 is a diagram for explaining an example of arrangement of reference pixels and encoded pixels.

[Explanation of symbols]

 100 predictive encoder 114 local decoder 116 line memory 120 predictor 130 quantizer 140 entropy encoder 150 inverse quantizer 151 adder 160 clamp circuit 170 memory 180 decoder 190 image processing circuit 192 printer

Claims (9)

[Claims] 1. Entropy coding after quantizing an error between a pixel value of a pixel to be coded and a predicted value generated based on a pixel previously coded and restored by a local decoder. Encoding means, and a clamp means for adjusting the dynamic range of the output of the local decoder so as not to exceed the dynamic range of the pixel values of the original image, and adjusting a clamp start point in the clamp means. The image encoding apparatus according to claim 1, wherein the background or foreground value decoded by the local decoder is controlled to be the same as the background or foreground value of the original image. 2. A predictive coding apparatus for predicting and coding a value of a pixel to be coded from a value of a reference pixel, and quantizing a prediction error, and coding an output of the quantizer. An entropy encoder, a local decoder that performs decoding based on the prediction error and the prediction value quantized by the quantizer, and when the output value of the local decoder reaches a predetermined threshold, Clamp means for clamping the output value to the maximum value or the minimum value of the pixel value of the original pixel, the predetermined threshold in the clamp means,
An image encoding device, wherein the pixel value of the original image is a value other than the maximum value or the minimum value. 3. The image coding apparatus according to claim 2, wherein said quantizer changes a level of a quantization error according to a value of a control parameter. 4. The control parameter is a value that defines a maximum width of an allowable quantization error, and the predetermined value in the clamp unit is determined based on a maximum value of a pixel value of an original image and the allowable quantum value. Value obtained by subtracting the value of the maximum width of
4. The image encoding apparatus according to claim 3, wherein the value is obtained by adding a value of a maximum width of the allowable quantization error to a minimum value of pixel values of an original image. 5. The image encoding device according to claim 1, wherein the image encoding device is a near lossless image encoding device.
The image encoding device according to any one of the above. 6. A signal coded by the image coding apparatus according to claim 2, which is decoded by performing a process reverse to the coding, and coding the decoded signal. An image decoding apparatus characterized in that a decoded signal is obtained by clamping under the same conditions as the clamping conditions used for the decoding. 7. The image encoding device according to claim 1, wherein the image encoding device comprises:
A memory for storing the encoded or decoded signal;
A communication device, comprising: an image processing device that performs predetermined image processing. 8. Entropy coding is performed after quantizing an error between a pixel value of a pixel to be coded and a predicted value generated based on a pixel previously coded and restored by a local decoder. An image encoding method, wherein when clamping the output signal of the local decoder so that the dynamic range of the output signal of the local decoder does not exceed the dynamic range of the pixel value of an original image, a clamp start point is set to an original value. An image encoding method, wherein the pixel value of an image is set to a value other than the maximum value or the minimum value. 9. In the quantization, a maximum width of an allowable quantization error is defined by a control parameter,
The clamp start point is defined as a value obtained by subtracting the maximum value of the allowable quantization error from the maximum value of the pixel value of the original image, or the minimum value of the pixel value of the original image. 9. The image encoding method according to claim 8, wherein a value obtained by adding the maximum width value is used.