JP2000350218A - Movement estimation device and method for image encoding system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a movement estimating device for image encoding system which increases the efficiency of image encoding by extending the motion estimation range of a P picture. SOLUTION: In the movement estimation device and method for image encoding system which increases the efficiency of image encoding by extending the movement estimation range of a P picture by using a reverse movement estimating device used for the movement estimation of a B picture when a picture type is a P picture, the picture type of an image to be transmitted is determined; when the picture type is a P picture, a forward movement estimation and backward motion estimation module divides an estimation area and carries out forward movement estimation. Then two obtained moving vectors (1st forward moving vector and 2nd forward movement vector) are compared and the moving vector having a smaller movement error is determined to reduce the movement error of the P picture, thereby increasing the encoding efficiency.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion estimation apparatus and method for extending a motion estimation range in an image coding system (for example, an MPEG-1 system or an MPEG-2 system). When estimating motion in a picture, an image in which the motion estimation range of a P-picture is extended to reduce motion errors and increase the efficiency of image encoding by using a backward motion estimator required for a B-picture. The present invention relates to a motion estimation device and method in a coding system.

[0002]

2. Description of the Related Art In general, MPEG-1 and MPEG-2, which are international standard proposals for moving picture compression, use composite motion compensation / compression.
The discrete cosine transform method is adopted. These standard proposals are all standard proposals for decoders, and the implementation of an encoder that satisfies these standards can be varied as long as decoding is possible. In the implementation of these image encoders, the portion with the largest load in terms of hardware (H / W) is the portion related to motion estimation. In particular, as the resolution of the standard input image increases, the configuration of the motion estimation part becomes more complicated, and the performance of the system is greatly affected by the method of implementing the motion estimation. That is, the complexity of hardware for motion estimation varies depending on the motion estimation range, and securing a wider estimation range is a very important part in terms of performance.

[0003] Normally, MPEG image compression methods include motion compensation and discrete cosine transform (DCT).
m) and variable length coding. Here, the motion compensation is
It depends on the image coding mode (picture type: I-picture, B-picture, P-picture) in MPEG.

That is, in the case of an I-picture, motion compensation is not required, and a P-picture can only be forward predicted from an I-picture or a P-picture previously coded. In addition, a B-picture can perform forward prediction and backward prediction from an I-picture or a P-picture coded earlier and later in time. Therefore, at the time of motion estimation, the encoder should obtain a motion vector (vector) from one search frame for a P-picture, and two types of motion vectors (a forward motion vector and a reverse motion vector) from two frames for a B-picture. ).

[0005] In general, a motion estimating unit is implemented by an ASIC (Application Specific Integrated System) for motion estimation.
Circuits), and in particular, for motion estimation of B-pictures, two modules for forward and backward motion vector estimation (a forward motion vector estimation module and a backward motion vector estimation module). ) Is required, and for a P-picture, only a forward motion vector estimation module is required, and a backward motion vector estimation module is not required.

FIG. 1 is a block diagram showing a configuration of a general MPEG encoder using composite motion compensation / discrete cosine transform. As shown in the figure, the motion estimator 1 extracts a motion vector by performing motion estimation using an input image and a previous image stored in a frame memory 2 for use in prediction. The motion compensating unit 3 uses a motion vector obtained from the motion estimating unit 1 to calculate a predicted value of an input image in units of macroblocks (16 × 16) currently processed in a reference image of a previous frame stored in the frame memory 2. To get from.

The predicted value and the input image value obtained in this way are subtracted by a subtractor 4 and the difference value is transmitted to a discrete cosine transform unit 5, and the discrete cosine transform unit 5 Is quantized, and the quantization unit 6 quantizes the DCT-processed data and outputs the resulting value to the variable-length encoding unit 7.
And the inverse quantization unit 8 respectively.

The variable length coding section 7 performs variable length coding (VLC: Variable Length Coding) on the quantized image data.
ding) and output as a bit string. Also,
The inverse quantization unit 8 inversely quantizes the image signal quantized by the quantization unit 6, and the inverse discrete cosine transform unit 9
The inversely quantized image signal is subjected to inverse DCT. In this manner, the inverse DCT image data is added by the adder 10 to the predicted value obtained from the motion compensation unit 3. The resulting image values in macroblock units are stored in the frame memory 2 as reference image values of the previous frame, and used for motion estimation and motion compensation of the next input frame. Here, the bit rate control unit 11
Controls the output of the quantization unit 6 by checking the buffer state in the variable length coding unit 7.

FIG. 2 is a diagram showing a conventional motion estimating device in an MPEG system, and FIG. 3 is a diagram for explaining a motion estimating mode in an MPEG system. First, in FIG. 3, motion estimation is not used for an I-picture, and motion estimation is performed for a P-picture using a previously encoded previous I-picture or P-picture as a search image. This is referred to as a directional motion vector, and such forward motion estimation is performed by the forward motion estimator 1a in FIG. Then, in FIG. 3, B-pictures B1 and B2 correspond to I-pictures before and after.
A picture or a P-picture is used as a search image to perform motion estimation. At this time, a motion vector estimated from a previous image is called a forward motion vector, and a motion vector estimated from an image thereafter is calculated in a backward direction. It is called a motion vector. Here, the backward motion estimation is performed by the backward motion estimation unit 1b in FIG.

Further, the forward motion vector and the backward motion vector obtained from the forward motion estimator 1a and the backward motion estimator 1b are transmitted to a motion vector processor 1c, and the motion vector processor 1c The motion vector is reconstructed (changed to a data format that can be processed at the rear end) and output according to the encoding mode (picture type). In FIG. 1, the motion vector output in this way is the motion compensator 3 and the variable-length encoder 7.
To be used for motion compensation of the input image block and for variable length encoding.

[0011]

In such an image system, compression of image data includes motion estimation and compensation,
It is performed through discrete cosine transform, variable-length coding, and the like. Among them, the motion estimator, in particular, is a relatively simple task, but requires enormous hardware.
Therefore, a typical image system uses an ASIC chip dedicated to motion estimation specially manufactured for this purpose.

However, the currently provided motion estimation AS
The IC is a chip in which the maximum search range of a motion vector is limited due to chip characteristics. Therefore, application systems that require a large search range (MPEG-1, MPEG-2, etc.)
In this case, since a plurality of these components are used, the burden on the hardware becomes heavy, and the disadvantage is that the motion estimation range must be limited particularly when the resolution of the image is very large, such as in high-definition television. Trigger.

The present invention has been proposed to solve the above-mentioned problems that occur in the motion estimator provided in the conventional image encoder. -By using the necessary backward motion estimation module for the picture,
It is an object of the present invention to provide a motion estimating apparatus and method in an image encoding system capable of extending a motion estimation range of a picture and improving the efficiency of image encoding.

[0014]

In order to achieve the above object, a motion estimating apparatus in an image encoding system according to the present invention provides an image encoding system for encoding and transmitting an image. A picture type discriminating means for discriminating a picture type, and performing motion estimation by comparing an original picture and an image locally decoded by an encoder according to the picture type discriminated by the picture type discriminating means; A first motion estimating means for extracting a forward motion vector as a reference, and performing a motion estimation by comparing an original image and an image locally decoded by an encoder according to the picture type determined by the picture type determining means. Second motion estimating means for extracting a backward motion vector or a forward motion vector as a result; A motion vector determining means for determining outputs of the two motion vectors respectively extracted by the first and second motion estimating means according to the determined picture type, and outputting only a motion vector corresponding to the determination result; And a motion vector processing means for reconstructing the motion vector output from the motion vector determination means and outputting it as a final motion vector.

Further, the first motion estimating means converts first and second scan formats of a decoded image locally decoded by an encoder and a scan format of an original image to be encoded for forward search. A two-scan format converter, and first and second demultiplexing the decoded image and the original image converted by the first and second scan format converters with an appropriate number of pixels to enable real-time processing. 2 data demultiplexer and the decoded image data obtained from the first data demultiplexer are stored in a first search data memory or a second search data memory.
A first memory controller for extracting and outputting the forward search data stored in the second search data memory or the first search data memory during motion estimation, and the second data demultiplexer; A second memory control unit for storing the original image data obtained in the reference data memory, extracting the data stored in the reference data memory during motion estimation and outputting the data as reference data,
And a motion estimator for extracting a forward motion vector after performing a motion estimation by comparing the reference data and the search data output from the second memory controller, respectively. is there.

The first memory control unit may simultaneously perform read and write operations of the first and second search data memories when the picture type is a P-picture and when the picture type is a B-picture. Are characterized by performing only a read operation from one of the two memories.

Further, the second motion estimating means converts a scan format of a decoded image locally decoded by an encoder and a scan format of an original image to be encoded for forward or backward search, respectively. A first and second scanning format converter, and a first for demultiplexing the decoded image and the original image respectively converted by the first and second scanning format converters with an appropriate number of pixels so that real-time processing can be performed. And a second data demultiplexer, and the decoded image data obtained from the first data demultiplexer is stored in a first search data memory or a second search data memory, and the second search data memory or the first search data is used for motion estimation. A first memory controller for extracting and outputting forward or backward search data stored in a memory, and an original image obtained from the second data demultiplexer; A second memory controller for storing data in a reference data memory, extracting the data stored in the reference data memory during motion estimation, and outputting the data as reference data, and output from the first and second memory controllers; A motion estimator for extracting a forward or backward motion vector after performing motion estimation by comparing the reference data and the search data with each other.

The first memory control unit may simultaneously perform read and write operations of the first and second search data memories when the picture type is a P-picture and when the picture type is a B-picture. Are characterized in that only one of the two memories performs a read operation.

When the picture type is a P-picture, the motion vector determining means compares the motion vectors obtained from the first and second motion estimating means, and determines, as a motion vector, one having a smaller error in the motion vector. If the picture type is a B-picture, the motion vectors obtained from the first and second motion estimating means are directly transmitted to the motion vector processing means.

When the picture type is a P-picture, the second motion estimating means performs motion estimation after receiving the forward search data, and extracts a forward motion vector as a result, When the picture type is B-picture, after receiving the backward search data, the apparatus performs motion estimation and extracts a backward motion vector based on the result.

Further, in the motion estimation method in the image encoding system for encoding and transmitting an image according to the present invention, the motion estimation method comprises the steps of: determining a picture type of an image to be transmitted;
In the case of a picture, a first motion vector extracting step of estimating a forward motion using a forward search image in a different region from the original image and extracting two forward motion vectors as a result, When the type is a P-picture, a first motion vector determining step of determining a smaller one of the two extracted forward motion vectors with a smaller error as a forward motion vector; and when the determined picture type is a B-picture, A second motion vector extraction step of performing forward motion estimation from the original image and the decoded search image to extract a forward motion vector, and, if the determined picture type is a B-picture, Extracting a third motion vector by performing a backward motion estimation from the decoded search image to extract a backward motion vector; When the type is a B-picture, a second motion vector determining step of determining the forward and backward motion vectors respectively extracted in the second and third motion vector extracting steps as a final motion vector. It is a feature.

[0022]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention based on the above-described technical concept will be described below in detail with reference to the accompanying drawings. FIG. 4 is a block diagram showing a motion estimating apparatus for extending a motion estimation range in an image coding system according to the present invention. Here, reference numeral 110 is
Reference numeral 120 denotes a picture type discriminator for discriminating a picture type of an image to be transmitted. Reference numeral 130 denotes a first motion estimator for extracting a forward motion vector.
10 illustrates a second motion estimator for extracting a backward motion vector or a forward motion vector by performing motion estimation from an original image and a decoded image according to the picture type determined in 10.

Also, reference numeral 140 determines the output of the two motion vectors extracted by the first and second motion estimators 120 and 130 according to the determined picture type, and corresponds to the determination result. A motion vector determining unit that outputs only the motion vector to be changed,
Reference numeral 50 denotes a motion vector processing unit that reconstructs the motion vector output by the motion vector determination unit 140 and outputs the reconstructed motion vector.

In the motion estimating apparatus according to the present invention, the picture type discriminating unit 110 first inputs the image block (macro block) to be encoded.
And the signals corresponding to the determined picture types are sent to the first and second motion estimators 120 and 130 and the motion vector determiner 140. introduce. Here, if the determined picture type is a P-picture, the first motion estimator 120 determines the reference data as the original image data and the forward search data locally decoded by the encoder for the search. To perform motion estimation, and as a result, extract a forward motion vector.

FIG. 5 is a diagram showing an embodiment of the first motion estimator 120. Here, the first scan format conversion unit 161 converts the scan format of the image data (current frame) locally decoded by the encoder for the motion estimation of the next frame into the vertical scan format. That is, the second scanning format converter 162 also converts the scanning format of the original image data (current frame) to be encoded into the scanning format in the vertical direction.

The first data demultiplexer 163 demultiplexes the decoded image data converted by the first scanning format converter 161 with an appropriate number of pixels so that real-time processing can be performed. 167. The first memory adjustment unit 167 stores the input search data in the first search data memory 165 or the second search data memory 166 and uses the search data as the motion estimation search data of the next frame. The search data of the previous frame stored in the search data memory 166 or the first search data memory 165 is extracted and transmitted to the motion estimator 170 for use in estimating the motion of the current frame.

Here, the first and second search data memories 1
Data reading and writing operations of 65 and 166 vary according to the picture type determined by the picture type determining unit 110. For example, if the picture type is P-
In the case of a picture, reading and writing of the first and second search data memories 165 and 166 are performed simultaneously,
When the picture type is a B-picture, only one of the two memories performs a read operation.

On the other hand, the second memory controller 169 stores the reference data of the current image obtained by the second data demultiplexer 164 in the reference data memory 168, and stores the reference data in a processing block (motion estimation is performed on a block basis). Therefore, only the reference data of the block is extracted) and extracted and transmitted to the motion estimator 170.

Since the picture type is a P-picture, the motion estimator 170 uses the first memory control unit 167 and the reference data as the image data of the current frame obtained by the second memory control unit 169. And performs motion estimation by comparing the search data of the previous frame with the search data of the previous frame. As a result, a first forward motion vector is extracted and transmitted to the motion vector determination unit 140.

As described above, when the picture type is a P-picture, in the past, only one forward motion vector was extracted by the forward motion estimator, but in the present invention, even when the picture type is a P-picture, In order to expand the motion estimation range, the second motion estimator 130 performs forward motion estimation by comparing search data in an area different from the search data of the first motion estimator 120 with the original image.

That is, the second motion estimating unit 130
As shown in (1), the first scan format conversion unit 181 converts the scan format of the image data (previous frame) locally decoded by the encoder into the vertical scan format, and The format conversion unit 182 also converts the scan format of the original image data (current frame) to be encoded into a vertical scan format.

The first data demultiplexer 183 demultiplexes the decoded image data converted by the first scanning format converter 181 with an appropriate number of pixels so that real-time processing can be performed. 187. The first memory controller 187 stores the input search data in the first search data memory 185 or the second search data memory 186 and uses the search data as motion search estimation data for the next frame. The search data of the previous frame stored in the search data memory 186 or the first search data memory 185 is extracted and transmitted to the motion estimator 190 for use in estimating the motion of the current frame.

Here, the first and second search data memories 1
Data read and write operations of 85 and 186 vary according to the picture type determined by the picture type determination unit 110. For example, since the picture type of the current input image is P-picture, reading and writing of the first and second search data memories 185 and 186 are performed simultaneously.

On the other hand, the second memory controller 189 stores the reference data of the current image obtained by the second data demultiplexer 184 in the reference data memory 188, and stores the reference data in a processing block (motion estimation is performed on a block basis). Therefore, only the reference data of the block is extracted) and extracted and transmitted to the motion estimator 190. The motion estimator 190
Since the picture type is a P-picture, reference data as image data of a current frame obtained by the second memory adjustment unit 189 and search data of a previous frame obtained by the first memory adjustment unit 187 , And performs motion estimation. As a result, a second forward motion vector is extracted and transmitted to the motion vector determining unit 140.

The motion vector determining unit 140 receives the first and second forward motion vectors as described above, and determines the P-picture according to the picture type determined by the picture type determining unit 110. In this case, one of the two forward motion vectors is selected and output as a forward motion vector. That is, if the picture type is a P-picture, the motion vector determination unit 140 may determine whether the first and second motion estimation units 120 and
The first and second forward motion vectors respectively obtained at 130 are compared, and a motion vector with a smaller motion compensation prediction error is selected, and the selected motion vector is used as a motion vector of a P-picture. To the unit 150.

The motion vector processing unit 150 reconstructs the input motion vector into a data format that can be processed at the rear end, and transmits the data format to the motion compensation unit 3 and the variable length coding unit 7 of FIG. And encoding are performed.

[0037]

As described above in detail, according to the motion estimating apparatus in the image coding system according to the present invention, when the picture type is P-picture, the search area is divided by the forward and backward motion estimating modules. Each of the two motion vectors is extracted by performing motion estimation, and the one having a smaller error among the two motion vectors is determined as the motion vector for the P-picture. Not only can the range be extended, but also the motion vector error can be minimized, and the efficiency of image coding can be increased.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a general MPEG encoder.

FIG. 2 is a diagram illustrating an embodiment of a conventional motion estimator applied to FIG. 1;

3A is a diagram illustrating a motion estimation mode of a P-picture, and FIG. 3A is a diagram illustrating a motion estimation mode of a P-picture, and FIG. 3B is a diagram illustrating a motion estimation mode of a B-picture. FIG.

FIG. 4 is a block diagram illustrating a motion estimation device in the image coding system according to the present invention.

FIG. 5 is a block diagram showing an embodiment of a first motion estimator of FIG. 4;

FIG. 6 is a block diagram illustrating an embodiment of a second motion estimator in FIG. 4;

[Explanation of symbols]

110 picture type discriminator, 120 forward motion estimator, 140 reverse motion estimator, 150 motion vector determiner, 160 motion vector processor.

Continuing on the front page (72) Inventor Kim ▲ Gwang ▼ Atsushi Modern Apartment 725-1804

Claims (8)

[Claims] 1. An image coding system for coding and transmitting an image, comprising: a picture type determining means for determining a picture type of an image to be transmitted; and an original image and an encoder based on the picture type determined by the picture type determining means. First motion estimating means for performing motion estimation by comparing the image with the locally decoded image and extracting a forward motion vector as a result; Second motion estimating means for performing motion estimation by comparing the image and the image locally decoded by the encoder, and thereby extracting a backward motion vector or a forward motion vector, The two motion vectors extracted by the first and second motion estimation means according to the picture type A motion vector determining means for determining a force and outputting only a motion vector corresponding to the determination result; and a motion vector for reconstructing a motion vector output from the motion vector determining means and outputting it as a final motion vector. A motion estimating apparatus in an image coding system, comprising: a processing unit. 2. The first and second motion estimating means comprises: a first and a second encoder for converting a scan format of a decoded image locally decoded by an encoder and a scan format of an original image to be encoded for forward search, respectively; A two-scan format converter, and first and second demultiplexing the decoded image and the original image converted by the first and second scan format converters with an appropriate number of pixels to enable real-time processing. The second image demultiplexer and the decoded image data obtained from the first data demultiplexer are stored in a first search data memory or a second search data memory, and the second image data is stored in a second search data memory.
A first memory controller for extracting and outputting the forward search data stored in the search data memory or the first search data memory, and storing the original image data obtained from the second data demultiplexer in the reference data memory; A second memory controller for extracting the data stored in the reference data memory and outputting the reference data as reference data when the motion is estimated, and comparing the reference data and the search data respectively output from the first and second memory controllers. 2. The motion estimating apparatus according to claim 1, further comprising: a motion estimator that extracts a forward motion vector after performing the motion estimation. 3. The first memory controller, if the picture type is a P-picture, performs the read and write operations of the first and second search data memories simultaneously, and if the picture type is a B-picture. 3. A motion estimating apparatus in an image encoding system according to claim 2, wherein the device performs only a read operation from one of two memories. 4. The second motion estimating means converts a scan format of a decoded image locally decoded by an encoder and a scan format of an original image to be encoded for forward or backward search, respectively. A first and second scanning format converter, and a first for demultiplexing the decoded image and the original image respectively converted by the first and second scanning format converters with an appropriate number of pixels so that real-time processing can be performed. And a second data demultiplexer, and the decoded image data obtained from the first data demultiplexer is stored in a first search data memory or a second search data memory, and the second search data memory or the first search data is used for motion estimation. A first memory controller for extracting and outputting forward or backward search data stored in a memory, and an original image data obtained from the second data demultiplexer; Stores the data in the reference data memory, a second output as the reference data pull the stored time to the reference data memory motion estimation data
A motion estimator for performing a motion estimation by comparing the reference data and the search data output from the first and second memory adjustment units with each other and extracting a forward or backward motion vector; The motion estimating apparatus in the image encoding system according to claim 1, comprising: 5. The first memory controller, if the picture type is a P-picture, performs the read and write operations of the first and second search data memories simultaneously, and if the picture type is a B-picture. 5. The apparatus according to claim 4, wherein only one of the two memories performs a read operation. 6. When the picture type is a P-picture, the motion vector determining means compares the motion vectors obtained from the first and second motion estimating means and determines a motion vector having a smaller error as a motion vector. 2. The image coding apparatus according to claim 1, wherein the motion vector obtained from the first and second motion estimating means is transmitted to the motion vector processing means as it is when the picture type is a B-picture. Motion estimation device in the system. 7. When the picture type is a P-picture, the second motion estimating means performs motion estimation after receiving forward search data, and extracts a forward motion vector as a result, 2. The image coding apparatus according to claim 1, wherein when the picture type is B-picture, after receiving backward search data, performing motion estimation and extracting a backward motion vector based on the result. Motion estimation device in the system. 8. A motion estimation method in an image encoding system for encoding and transmitting an image, comprising: a picture type determining step of determining a picture type of an image to be transmitted, wherein the determined picture type is a P-picture. A first motion vector extracting step of estimating a forward motion using a forward search image of a different region from the original image and extracting two forward motion vectors as a result, wherein the picture type is P-picture In the case of the first motion vector, a first motion vector determining step of determining, as the forward motion vector, the one with the smaller error among the two extracted forward motion vectors; and, if the determined picture type is a B-picture, Second motion vector extraction for performing forward motion estimation from the searched image and extracting a forward motion vector. A third motion vector extracting step of performing a backward motion estimation from the original image and the decoded search image to extract a backward motion vector when the determined picture type is a B-picture; When the picture type is a B-picture, a second motion vector determining step of determining the forward and backward motion vectors respectively extracted in the second and third motion vector extracting steps as a final motion vector. A motion estimation method in an image coding system, characterized in that: