KR0178223B1 - Improved pattern vector encoding system and codeword renewal method thereof - Google Patents

Abstract

The present invention provides an improved pattern vector encoding system and a codeword used for updating a codeword of a reference pattern corresponding to an input pattern corresponding to a divided subblock for pattern vector encoding according to a frequency of use thereof. According to the present invention, a memory includes a first memory area for storing preset codewords and a second memory area for storing new codewords to be updated. The codewords of the respective reference patterns corresponding to the respective input patterns for are searched in the first memory area to determine optimal reference patterns having the least error value for each input pattern, and then for each of the determined reference patterns. Index values are provided to the vector quantization means, and the squared absolute difference between each input pattern and the reference pattern is checked. Pattern matching means for generating an update signal when the code word for the update exceeds a predetermined drain shawl deugap; Third memory means for temporarily storing a new codeword to be updated corresponding to the corresponding input pattern generated based on the update signal from the pattern matching means; A pointer for storing an address of a new codeword stored in a predetermined portion of said third memory means; Counting means for counting the number of occurrences of the new codeword by incrementing the reference for the new codeword based on the update signal from the pattern matching means, and the pattern matching means further comprising: the number of occurrences of the new codeword. When a predetermined number of times is a predetermined number of times, a new codeword stored in the third memory means is read out, stored in the second memory area, and provided to the output side. Each index value from the vector quantization means and codewords from the pattern matching means are read. It includes a multiplexer for multiplexing and transmitting.

Therefore, according to the present invention, by performing vector quantization while continuously updating codewords according to an input pattern corresponding to each subblock divided into a predetermined size, it is possible to effectively improve image quality deterioration of the reconstructed image as well as efficient quantization. It is.

Description

Improved Pattern Vector Coding System and Its Codeword Update Method

1 is a block diagram of an improved pattern vector coding system according to a preferred embodiment of the present invention.

2 is a block diagram of an improved pattern vector encoding system according to an embodiment of the present invention.

3 is a detailed block diagram of a codeword memory block employed in the improved pattern vector encoding system of the present invention.

4 is a flowchart illustrating a process of updating a codeword to adapt to a pattern of an input image according to the present invention.

5 is a block diagram of a conventional typical pattern vector coding system.

6 is a diagram illustrating an example of dividing one DCT block into four subblocks or 16 subblocks as an example in the improved pattern vector coding system or the conventional pattern vector coding system according to the present invention.

* Explanation of symbols for main parts of the drawings

1030 frame memory 12 subtractor

14 DCT block 16 VQ block

18: control block 20: memory block

22: counter 24: block recovery block

26: IDCT block 28: adder

32: motion prediction block 34: multiplexer

202 and 204 memory 2041 address processing block

2043: buffer 2045: pointer

2047: reference processing block

The present invention relates to an image coding system using pattern vector quantization. More particularly, the present invention relates to an improved pattern vector capable of updating codewords to adapt to various image changes in an encoding system performing pattern vector quantization using codewords. It relates to an encoding system.

As is well known in the art, the transmission of discrete video signals can maintain better image quality than analog signals. When a video signal composed of a series of image frames is represented in digital form, a considerable amount of data must be transmitted, especially for high-definition television (HDTV). However, since the usable frequency range of the conventional transmission channel is limited, in order to transmit a large amount of digital data, it is necessary to compress the data to be transmitted and reduce the amount of transmission. In addition, the compressed video signal and the audio signal are encoded through different encoding techniques due to the characteristics of the signals. In this encoding, a compression technique of a video signal in which a larger amount of digital data is generated than an audio signal is particularly important. It can be said to take part.

On the other hand, as the various compression methods mainly used for encoding the video signal, the hybrid coding method combining the stochastic coding method and the temporal and spatial compression method is known to be the most efficient.

Most hybrid coding schemes, which are one of the efficient coding schemes described above, use motion compensated DPCM (differential pulse code modulation), two-dimensional discrete cosine transform (DCT), quantization of DCT coefficients, VLC (variable length coding), and the like. Here, the motion compensation DPCM determines a motion of the object between the current frame and the previous frame, and predicts the current frame according to the motion of the object to generate a differential signal representing the difference between the current frame and the prediction value. These methods are, for example, Staffan Ericsson's Fixed and Adaptive Predictors for Hybrid Predictive / Transform Coding, IEEE Transactions on Communication, COM-33, NO.12 (1985, December), or A motion Compensated Interframe from Ninomiy and Ohtsuka. Coding Scheme for Television Pictures, IEEE Transactions on Communication, COM-30, NO.1 (January, 1982).

More specifically, the motion compensation DPCM predicts the current frame from the previous frame according to the motion of the object estimated between the current frame and the previous frame. Here, the estimated motion may be represented by a two-dimensional motion vector representing the displacement between the previous frame and the current frame.

In general, there are several approaches to estimating the pixel displacement of an object, and they are generally classified into two types, one of which is a block-based motion estimation method and the other is a pixel-based motion estimation method. In motion estimation, the block of the current frame is compared with the blocks of the previous frame to determine an optimal matching block. From this, the interframe displacement vector (how much the block has moved between frames) with respect to the entire frame is transmitted. It is estimated.

Therefore, when transmitting a video signal, the transmitting side compresses and encodes the image signal in the buffer of the output side in order by taking into account the spatial and temporal correlation of the image signal in block units or pixel units through the above-described encoding technique. The video data is transmitted to the decoding system on the receiving side through the transmission channel at a desired bit rate in response to the channel request.

More specifically, the encoding system on the transmitting side removes spatial redundancy of the video signal by using transform coding such as discrete cosine transform (DCT), and further uses temporal encoding of the video signal by using differential coding through motion estimation and prediction. By eliminating redundant redundancy, the video signal can be efficiently compressed.

However, the DPCM / DCT hybrid coding technique described above has a target bitrate of Mbps, and its application fields are CD-ROM, computers, home appliances (digital VCRs, etc.), broadcasting (HDTV), and the like. It is mainly concerned with MPEG-1, 2 and H.261 encoding algorithms for high bit rate encoding, which mainly consider only the statistical characteristics of block-by-block motion in the image, which has already been completed by the standard.

On the other hand, in recent years, due to the improvement and spread of PC performance, the development of digital transmission technology, the realization of high-definition display device, the development of memory device, various devices such as home appliances can process and provide image information with huge data. In order to meet this demand, the transmission of audio-video data through the existing low-speed transmission path (PSTN, LAN, mobile network, etc.) and the storage capacity of limited capacity to meet these demands. New storage techniques with high compression rates are needed for storage.

However, as described above, the conventional video coding system that removes temporal and spatial redundancy of video signals using MC-DCT (motion compensation DCT) can obtain some characteristics in terms of coding efficiency. Although it is possible, in practical implementation, there is a problem in that it is difficult to apply to a system requiring a high compression ratio as described above because of the limitation of the compression ratio.

Therefore, in consideration of the above-described points, instead of the hybrid technique using MC-DCT, the DCT block is divided into subblocks of a predetermined size as a technique capable of satisfying the high compression ratio, and then a code table having a plurality of codewords is used. By applying vector quantization, a pattern vector coding system for performing pattern vector coding has been proposed. An example of such a typical conventional pattern vector coding system has a form as shown in FIG.

As shown in the figure, the conventional pattern vector coding system includes a frame memory 100, a discrete cosine transform block (hereinafter abbreviated as DCT) 110, and a vector quantization (VQ) block 120. , Pattern matching block 130 and code table 140.

In FIG. 4, the DCT block 110 uses the cosine function of the video signal (pixel data) in the time domain provided from the frame memory 100 on the input side of the DCT block 110 in a frequency domain of 8 × 8 unit blocks. Convert to

Then, the DCT transform coefficients transformed in 8 × 8 block units are provided to the next vector quantization block 120. Accordingly, in the vector quantization block 120, for typical pattern vector coding, the DCT transform coefficients of the 8x8 block as shown in FIG. 5 (a) are further divided into subblocks of a predetermined size, e.g. As shown in Fig. 5 (b), each 8 × 8 size DCT transform coefficient block is divided into four subblocks of 4 × 4 size, or as shown in Fig. 5 (c). It is divided into 16 subblocks of 2 sizes. Here, an example of dividing each DCT block into four subblocks of 4x4 will be described.

That is, in the vector quantization block 120, each subblock (for example, four subblocks for one DCT block) obtained by dividing each DCT change coefficient block provided from the DCT block 110 is divided into pattern matching blocks. To 130.

Accordingly, the pattern matching block 130 searches for the codeword (reference pattern) having the value most similar to the input subblocks provided from the vector quantization block 120, that is, the input pattern, from the code table 140. A reference pattern is determined, that is, an optimal reference pattern having the smallest error value is determined for the corresponding input pattern corresponding to the subblock, and then the index value for the determined reference pattern is converted into the vector quantization block 120. As a result, the vector quantization block 120 transmits index values corresponding to each input pattern to a transmitter not shown.

That is, in the conventional pattern vector coding system, as in the conventional hybrid coding scheme, by using the MC-DCT, the video signal is not compressed and encoded by removing the temporal and spatial redundancy of the image. By dividing the DCT blocks into four 4 × 4 subblocks, and then performing pattern vector encoding using codewords on each of the divided subblocks, an effective high compression rate image encoding is implemented.

However, in the conventional pattern vector coding system as described above, if the corresponding subblock does not match the codeword in the code table by searching for a pattern corresponding to each subblock divided in a limited codeword, As a result, vector quantization is not performed, resulting in deterioration of the image quality finally restored in the decoding system at the receiving side, or when there is no codeword that can actually correspond to the divided subblock, effective pattern search is difficult. It is becoming a deterrent to high speed realization for real time processing under compression ratio.

Accordingly, the present invention is to solve the above problems of the prior art, an improved pattern capable of updating the codeword of the reference pattern corresponding to the input pattern corresponding to the sub-block divided for the pattern vector encoding according to the frequency of use The purpose is to provide a vector encoding system.

Another object of the present invention is to provide an improved pattern vector encoding system capable of updating a codeword of a reference pattern corresponding to an input pattern corresponding to a divided subblock for high-compression pattern vector encoding according to a frequency of use thereof. .

Another object of the present invention is to provide a codeword updating method capable of adaptively updating codewords for a pattern of each input image in pattern vector encoding.

In accordance with one aspect of the present invention, there is provided a plurality of subblocks of a predetermined size of 8 × 8 DCT transform coefficient blocks obtained through a discrete cosine transform using a cosine function for each input image. A pattern for dividing and outputting indexes corresponding to the input patterns of the respective subblocks by performing quantization through vector quantization means using a plurality of predetermined codewords provided in the memory for each of the divided subblocks. In the vector encoding system, the memory includes a first memory area for storing the predetermined codewords and a second memory area for storing new codewords to be updated, each of the divided subblocks. The codewords of each of the reference patterns corresponding to the input pattern are searched in the first memory area, and the The optimum reference patterns having the least error value are determined for the output pattern, and then index values for the determined respective reference patterns are provided to the vector quantization means, and the squared absolute difference between each input pattern and the reference pattern is determined. Pattern matching means for generating an update signal for updating codewords when it checks and exceeds a predetermined threshold value; Third memory means for temporarily storing a new codeword to be updated corresponding to the corresponding input pattern generated based on the update signal from the pattern matching means; A pointer for storing an address of the new codeword stored in a predetermined portion of the third memory means; Counting means for counting the number of occurrences of the new codeword by incrementing a reference for the new codeword based on an update signal from the pattern matching means, wherein the pattern matching means includes: When the number of occurrences is a predetermined number of times, the new codeword stored in the third memory means is read out, stored in the second memory area, and provided to the output side, and the pattern matching is performed with the respective index values from the vector quantization means. And a multiplexer for multiplexing and transmitting the respective codewords from the means.

According to another aspect of the present invention, there is provided a discrete cosine using a cosine function for a difference signal between a current input frame and a prediction frame generated through motion estimation compensation between the current frame and a previous frame. Vector quantization means for dividing each 8 × 8 DCT transform coefficient block obtained through the transformation into a plurality of subblocks of a predetermined size, and using a plurality of predetermined codewords included in the memory for each of the divided subblocks. In the pattern vector encoding system for outputting an index corresponding to the input patterns of each subblock by performing quantization, the memory includes a first memory area for storing the predetermined codewords and a new code to be updated. And a second memory area for storing words, each subblock corresponding to the divided subblocks. Codewords of each of the reference patterns corresponding to each input pattern are searched in the first memory area to determine optimal reference patterns having the least error value for each of the input patterns, and then to each of the determined reference patterns. Pattern matching means for providing the index values to the vector quantization means, and checking an absolute square difference between the input pattern and the reference pattern and generating an update signal for updating a codeword when a predetermined threshold value is exceeded; Third memory means for temporarily storing a new codeword to be updated corresponding to a corresponding input pattern generated based on the update signal from the pattern matching means; A pointer for storing an address of the new codeword stored in a predetermined portion of the third memory means; Counting means for counting the number of occurrences of the new codeword by incrementing a reference for the new codeword based on an update signal from the pattern matching means, wherein the pattern matching means includes: When the number of occurrences is a predetermined number of times, the new codeword stored in the third memory means is read out, stored in the second memory area, and provided to the output side. The index values and the index values provided from the vector quantization means are provided. An improved pattern vector encoding system comprising a multiplexer for multiplexing and transmitting the respective codewords from the pattern matching means.

According to another aspect of the present invention, the 8 × 8 DCT transform coefficient blocks obtained through discrete cosine transform are divided into a plurality of subblocks of a predetermined size, and for each of the divided subblocks. A method of updating a codeword in a system for outputting index values corresponding to input patterns of each subblock by performing pattern vector encoding using a plurality of preset codewords included in a memory, the memory may include: A first memory area for storing a plurality of preset codewords and a second memory area for storing new codewords to be updated, and when a sub-block input pattern is applied, a reference pattern from the first memory area Retrieving codewords of to determine an optimal reference pattern having the smallest error value for the corresponding input pattern; Calculating a squared absolute difference value between the corresponding input pattern and the reference pattern, and checking whether the calculated absolute difference value exceeds a predetermined threshold value; If it is determined that the calculated absolute difference value exceeds the preset threshold value, the new codeword corresponding to the input pattern is stored in a temporary buffer and the corresponding count value is increased by one. step; If the count value for the new codeword is checked and the value is not a predetermined count value, the process is repeated. If the checked count value and the preset count value match, the count value stored in the temporary buffer is stored. Reading a new codeword and storing it as an updated new codeword in the second memory area; A method of updating a codeword in a pattern vector encoding system comprising multiplexing the updated new codeword with an index value corresponding to the input patterns of each subblock is output.

The above and other objects and various advantages of the present invention will become more apparent from the preferred embodiments of the present invention described below with reference to the accompanying drawings by those skilled in the art.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

First, the pattern vector encoding to be used in the present invention has a code table, and the reference patterns (codewords) in the code table are divided into regions having a predetermined size, for example, 4 × 4 or 2 × 2, of the input image. The most technical feature is that vector quantization is performed continuously according to an input pattern corresponding to a block (i.e., a new input pattern generated due to a change in image). When a predetermined number of times or more continue, the codeword corresponding to the input pattern is recorded in the memory and used as one of the codewords of the corresponding input pattern to be input again later.

Of course, according to the present invention, the newly updated codewords are transmitted to the receiving end of the remote place through the transmitter together with the pattern vector coded indices. Therefore, the pattern vector coded image data at the receiving side may be restored to the original signal through the decoding system at the receiving side having the codewords updated in the same manner as the encoding system at the transmitting side.

Example 1

1 shows a schematic block diagram of an improved pattern vector coding system according to a first embodiment of the present invention. As can be seen from the figure, the video encoding system according to the present invention comprises a control block 18, a memory block 20, and a counter 22 as pattern matching means for pattern vector encoding. Having the main features of the phase, the objects of the present invention are achieved by these components.

In FIG. 1, each of the frame memory 10, the DCT block 14, and the vector quantization block 16 are the same components that perform substantially the same functions as those in FIG. Since the operation of these components is already described in detail above, the description thereof will be omitted to avoid unnecessary duplication.

Therefore, if 8 × 8 DCT transform coefficient blocks are provided from the DCT block 14, the vector quantization block 16, as in the above-described prior art, is shown in Fig. 5A for pattern vector encoding. The 8 × 8 DCT transform coefficient blocks are divided into subblocks of a predetermined size, for example, each 8 × 8 DCT transform coefficient block is divided into 4 × as shown in FIG. It is divided into four subblocks of 4 sizes or into 16 subblocks of 2x2 size, as shown in FIG. In this embodiment, as shown in FIG. 5C, the case where each DCT block is divided into four sub-blocks of 4x4 will be described as an example.

That is, the vector quantization block 16 controls each subblock (eg, four subblocks for one DCT block) obtained by dividing each DCT change coefficient block provided from the DCT block 14 at the front end. Block 18 is provided.

Accordingly, the control block 18 performs pattern matching with the subblocks input from the vector quantization block 16, that is, the input patterns, and the reference patterns preset in the memory block 20. . That is, the control block 18 searches the memory block 20 for a codeword (reference pattern) having a value most similar to the input pattern to determine the reference pattern, that is, the error value for the corresponding input pattern corresponding to the subblock. After determining the least optimal reference pattern, an index value for the determined reference pattern is output to the vector quantization block 16 and output to the multiplexer (MUX) 34 on the output side.

At this time, the control block 18 checks whether or not the absolute value of the square of the reference pattern (codeword) having the most similar value to the corresponding input pattern and the input pattern exceeds the predetermined threshold value as shown in the following equation. If it is determined that the absolute difference value of the square between the reference pattern and the input pattern exceeds the preset threshold value, a control signal for updating to a new codeword corresponding to the input pattern is generated and a new codeword is generated. Is temporarily stored in the latch, and then the counter value of the counter 22 is set to one.

| Vin-Cin | ² Th

In the above formula, Vin denotes an input pattern, Cin denotes a reference pattern, and Th denotes a preset threshold value.

More specifically, the memory block 20 may include a first memory 202 in which codewords for reference patterns are stored and a second memory for updating a new codeword, as shown in FIG. 204, wherein the second memory 204 for updating a new codeword is composed of an address processing block 2041, a buffer 2043, and a latch, as shown in FIG. A pointer 2045, and a reference processing block 2047.

In FIG. 3 (b), the address processing block 2041 manages new codewords to be updated stored in the buffer 2043, and also performs processing such as updating of new codewords. When a codeword is generated a predetermined number of times and is to be stored in the first memory 202 as a reference pattern, the corresponding codeword is read from the buffer 2043, and the read codeword is read again through the line L13. Provide to 18.

In addition, when an update signal for a new codeword is input from the control block 18 through the line L11, the address processing block 2041 writes this new codeword into the buffer 2043, and then latches the address of the new codeword. To the pointer 2045. At this time, the reference processing block 2047 substantially consists of an adder and a memory, and processes the reference count value. That is, when the update signal of the new codeword is input from the control block 18 through the line L12, the reference processing block 2047 increments the corresponding reference for codeword update by 1 and then increases the count value of the counter 22 by one. Let's do it. Here, the buffer 2043 is for temporarily storing new codewords to be updated as a first-in first-out buffer.

When new codewords to be updated are stored in the buffer 2043 through the above-described process, the count value for each new codeword is checked to determine whether the count value has been a predetermined number of times, for example, 10 times. If it is determined that the count value of the corresponding codeword to be updated is 10 times (that is, a new codeword has occurred 10 times in a predetermined time), the address processing block 2041 is buffer 2043. The corresponding new codeword is read from and provided to the control block 18 of FIG. 3 (a) via line L13. At this time, when a new codeword to be updated is generated 10 times and provided to the control block 18, the new codeword stored in the buffer 2043 is deleted at the same time, and the corresponding count in the counter 22 is deleted. The value is cleared.

That is, the control block 18 stores a new codeword corresponding to the input pattern read out from the buffer 2043 in the first memory 202 as a reference pattern, and stores an index value for the new codeword. A new codeword is output to the multiplexer (MUX) 34 of FIG. 1 while being provided to the vector quantization block 16 of FIG. In this case, the codewords stored in the first memory 202 are sequentially recorded in the order of input, and when the capacity is filled up, the limit of the memory capacity may be overcome by sequentially discarding the first input codeword values. will be. Here, the first memory 202 has a separate area for storing the existing codeword and the new codeword to be updated, and only the memory area for the new codeword serves as a first-in first-out buffer.

Therefore, the multiplexer 34 multiplexes the new codewords provided from the control block 18 and the index values provided from the vector quantization block 16 and transmits them to a transmitter not shown.

4 is a flowchart illustrating a process of retrieving and outputting a codeword corresponding to each input pattern or updating to a new codeword according to the present invention. Therefore, the following describes the process of updating a new codeword in more detail with reference to the flowchart of FIG. 4 mainly.

Referring to FIG. 4, when an input pattern is input from the vector quantization block 16 (step S10), the control block 18 searches for an area of the first memory 202 of FIG. The reference pattern (codeword) having the value most similar to the pattern is read (step S11).

Then, the control block 18 checks whether the square absolute difference value between the reference pattern (codeword) and the corresponding input pattern is larger than the preset threshold value (step S12), and then the square absolute difference value is preset. If it is determined that the threshold value is smaller than the threshold value, the process proceeds to step S19 described later, so that the control block 18 outputs the index value for the codeword corresponding to the corresponding input pattern to the vector quantization block 16. do.

On the other hand, if it is determined in the step S13 that the absolute absolute difference value is larger than the preset threshold value, the control block 18 controls to update to a new codeword that can correspond to the corresponding input pattern. Generates a signal and temporarily stores a new codeword corresponding to the corresponding input pattern in the buffer 2043 of FIG. 3B (step S13), and then increments the count value of the counter 22 by one (step S14). ).

Then, the number of counts for the new codeword to be updated counted at the counter 22 is checked, i.e. whether the count value for the new codeword is a predetermined number N, for example, 10 times. Is checked (step S15), if it is determined that N is not equal to 10, the process proceeds to step S10 described above to repeat the subsequent steps.

On the other hand, if it is determined that the count value for the new codeword is N, which is a predetermined number of times, as a result of the check in the step S15, the address processing block (based on the read address signal from the control block 18) 2041 reads the new codeword value from the buffer 2043 and provides it to the control block 18 again (step S16), so that the control block 18 sends the new codeword to the first memory 202. Save (step S17).

Therefore, the update of the new codeword generated a predetermined number of times through this process is performed. In this case, the respective codewords stored in the first memory 202 are sequentially recorded in the order of input. When the capacity is full, the first codewords 202 are discarded sequentially from the inputted codeword values. That is, when the capacity of the first memory 202 is completely filled, the frequency of occurrence is low and the old codeword values are discarded.

Therefore, by providing the index values for the codewords corresponding to the respective input patterns to the vector quantization block 16 at the control block 18 (steps S18 and S19), the vector quantization block 16 at the DCT block. Each index value corresponding to each input pattern of sub-blocks divided into predetermined sizes provided from 14 is output to the next multiplexer 34. Of course, in the control block 18, when the codeword of the index value provided to the vector quantization block 16 is an updated new codeword, the updated codeword is provided to the multiplexer 34 on the output side.

The multiplexer 34 then multiplexes the updated codewords provided from the control block 18 and the index values from the vector quantization block 16 and transmits them to the transmitter, not shown, thereby providing a desired method according to the present invention. The pattern vector coded video signal is sent to the receiving side of the remote site.

Therefore, according to the present embodiment, the codewords are continuously updated according to an input pattern or a new input pattern (that is, a new codeword generated due to a change in image) corresponding to each subblock divided into a predetermined size. By performing the vector quantization, not only efficient quantization but also image quality deterioration of the received reconstructed image can be effectively improved.

Example 2

2 shows a schematic block diagram of an improved pattern vector coding system according to a second embodiment of the present invention.

This embodiment is substantially the same in pattern vector encoding as in the first embodiment described above, in that pattern vector encoding is performed while updating not only a predetermined codeword but also a new codeword that can be generated according to a change in an image. However, the most distinctive difference is that the pattern vector coding can be performed while updating the codeword for the motion compensated differentially coded video frame obtained by employing the MC-DCT. Therefore, the present embodiment may enable high compression of the input image as well as the effect obtained in the above-described first embodiment.

Referring to FIG. 2, in the present embodiment, the same reference numerals are shown as like reference numerals in order to facilitate understanding of the same constituent members which perform substantially the same functions as in FIG. . Therefore, in the following, since the functions of the other components have already been described in detail above, the present invention will be described based on the operation of the newly added components for motion estimation compensation in consideration of the time axis of the image.

First, local reconstruction and prediction means for generating a predictive frame for motion estimation compensation in consideration of the time axis of an image according to the present embodiment, as is apparent from FIG. 2, includes a block reconstruction block 24, an IDCT block 26, The adder 28 includes a second frame memory 30, a motion prediction block 32, and a subtractor 12.

In FIG. 2, the block reconstruction block 24 divides each DCT block provided from the vector quantization block 16 into four 4x4 subblocks, respectively, and each of the divided subblocks. For the index values for the reference patterns coded with the pattern vector for, the 8 × before the pattern vector is coded using the control block 18 and the memory block 20 in which the existing codeword and the updated new codeword are stored. Restore to 8 DCT transform coefficient blocks.

In this embodiment, although the same control block 18 and the memory block 20 are shown to be used jointly at the time of vector quantization and block reconstruction for motion estimation compensation, when searching for the reference pattern with respect to the input pattern Each may be configured separately in consideration of the possibility of collision or waiting (priority setting) that may be caused.

Accordingly, in the IDCT block 26, the DCT transform coefficients of the reconstructed image frame provided from the block reconstruction block 24 are used as original signals (pixel data) before transform-coding for motion estimation and compensation. It restores and provides to the adder 28 of the next stage.

Then, the adder 28 adds the restored current frame signal (differential signal) provided from the IDCT block 26 and the predicted frame signal provided from the motion side block 32 through the line L22 to add a second frame. By providing the memory 30, the second frame memory 30 is stored as a frame immediately before the restored current frame signal, that is, the current frame signal input for current encoding.

Therefore, through this process, the previous frame signal stored in the second frame memory 30 is continuously updated as the image data immediately before the currently encoded input image data. The previous frame signal updated in this way is provided to the motion prediction block 32 of the next stage for motion prediction and compensation, and temporal redundancy between successive frames is removed through this series of processes.

That is, in the motion prediction block 32, the current frame from the first frame memory 10 and the previous frame from the second frame memory 30 are within a predetermined search range, for example, in units of 16 × 16. The motion is estimated, i.e., the block of the previous frame most similar to the block of the current frame is determined, and then the determined corresponding block is inputted to the above-described subtractor 12 and adder 28 through LINY L22.

In addition, although not shown in FIG. 2, a series of motion vectors determined by the motion prediction block 32 are encoded through a predetermined encoding process for transmission to a receiving side decoding system, and then sent to a transmitter (not shown). .

As a result, the subtractor 12 subtracts the difference signal (differential pixel) by subtracting the current frame signal from the first frame memory 10 with the motion estimation provided from the motion prediction block 32 and the compensated prediction frame signal. Value), and the difference signal between the current frame and the motion estimation compensated prediction frame is compressed and coded at a predetermined compression rate through DCT and vector quantization as described above.

In addition, the motion prediction block 32 provides the multiplexer 34 on the output side with a series of motion vectors generated in estimating the motion between the current frame and the previous frame, although the illustration in FIG. 2 is omitted. ), Multiplexed the index values from the vector quantization block 16, the updated new codewords from the control block 18, and the series of motion vectors from the motion prediction block 32 to a transmitter not shown. Will be sent.

Therefore, the present embodiment performs pattern vector coding while updating codewords on a motion compensated differentially coded image frame obtained by employing MC-DCT, thereby achieving the effect obtained in the above-described first embodiment as well as effective high compression ratio. Pattern vector coding can be realized.

Claims (16)

Each 8 × 8 DCT transform coefficient block obtained through a discrete cosine transform using a cosine function for each input image is divided into a plurality of subblocks of a predetermined size, and a plurality of subblocks provided in the memory for each of the divided subblocks. In the pattern vector encoding system for outputting an index corresponding to the input patterns of each subblock by performing quantization by vector quantization means using predetermined codewords of the memory, the memory stores the preset codeword. A first memory area for storing the second code area and a second memory area for storing new codewords to be updated, wherein the codewords of the respective reference patterns corresponding to the respective input patterns of the divided subblocks are stored in the first memory area; Search the memory area to find the optimal reference patterns with the lowest error value for each input pattern. After the determination, the index values for the determined reference patterns are provided to the vector quantization means. The absolute value of the square between the respective input patterns and the reference pattern is checked to update the codeword when the threshold value is exceeded. Pattern matching means for generating a signal; Third memory means for temporarily storing a new codeword to be updated corresponding to a corresponding input pattern generated based on the update signal from the pattern matching means; A pointer for storing an address of the new codeword stored in a predetermined portion of the third memory means; Counting means for counting the number of occurrences of the new codeword by incrementing a reference for the new codeword based on an update signal from the pattern matching means, wherein the pattern matching means includes: When the number of occurrences is a predetermined number of times, the new codeword stored in the third memory means is read out and stored in the second memory area and provided to the output side, and the index values and the pattern from the vector quantization means are provided to the output side. And a multiplexer for multiplexing and transmitting the respective codewords from the matching means. 2. The improved pattern vector encoding system of claim 1, wherein the plurality of divided subblocks are four subblocks obtained by dividing each of the 8x8 DCT transform coefficient blocks into 4x4 units. 2. The improved pattern vector encoding system of claim 1, wherein the plurality of divided subblocks are 16 subblocks obtained by dividing each of the 8x8 DCT transform coefficient blocks into 2x2 units. 4. The improved pattern vector encoding system of claim 1, 2 or 3, wherein the second memory area is a first-in first-out buffer memory that sequentially erases the new codewords according to an input order thereof. 4. The improved pattern vector coding system of claim 1, 2 or 3, wherein the pointer is comprised of a latch. Each 8 × 8 DCT transform coefficient block obtained through a discrete cosine transform using a cosine function for the difference signal between the current input frame and the prediction frame generated through the motion estimation compensation between the current frame and the previous frame is obtained. By subdividing into a plurality of subblocks and performing quantization through vector quantization means using a plurality of predetermined codewords provided in the memory for each of the divided subblocks, the subblocks correspond to the input patterns of the subblocks. In the pattern vector encoding system for outputting an index, the memory includes a first memory area for storing the predetermined codewords and a second memory area for storing new codewords to be updated, the divided angles Codewords of the respective reference patterns corresponding to each input pattern for the subblock Search in the first memory area to determine optimal reference patterns having the lowest error value for each input pattern, and then provide index values for the determined reference patterns to the vector quantization means, wherein each input pattern is obtained. Pattern matching means for checking an absolute difference between a pattern and a reference pattern and generating an update signal for updating codewords when a predetermined threshold value is exceeded; Third memory means for temporarily storing a new codeword to be updated corresponding to a corresponding input pattern generated based on the update signal from the pattern matching means; A pointer for storing an address of the new codeword stored in a predetermined portion of the third memory means; Counting means for counting the number of occurrences of the new codeword by incrementing a reference for the new codeword based on an update signal from the pattern matching means, wherein the pattern matching means generates the new codeword. When the number of times is a predetermined number of times, the new codeword stored in the third memory means is read out and stored in the second memory area and provided to the output side, and the index values and the patterns provided from the vector quantization means are provided. And a multiplexer for multiplexing and transmitting the respective codewords from the matching means. 7. The improved pattern vector encoding system of claim 6, wherein the plurality of divided subblocks are four subblocks obtained by dividing each of the 8x8 DCT transform coefficient blocks by 4x4 units. 7. The improved pattern vector encoding system of claim 6, wherein the plurality of divided subblocks are 16 subblocks obtained by dividing each of the 8x8 DCT transform coefficient blocks by 2x2 units. 9. The improved pattern vector encoding system of claim 6, 7, or 8, wherein the second memory area is a first-in first-out buffer memory that sequentially erases the new codewords according to an input order thereof. 9. The improved pattern vector coding system of claim 6, 7, or 8, wherein said pointer is comprised of a latch. The method of claim 6, 7, or 8, wherein the encoding system performs pattern vector encoding on the divided subblocks using the same pattern matching means, and generates the prediction frame. And inversely quantize index values for each input pattern of the sub-blocks encoded in the pattern vector and perform inverse discrete cosine transform. 10. The apparatus of claim 6, 7, or 8, wherein the encoding system comprises: the pattern matching means for performing pattern vector encoding on the plurality of divided subblocks, and the pattern vector for generating the prediction frame. And other pattern matching means for inversely quantizing index values for each input pattern of the encoded subblocks and performing inverse discrete cosine transform. Each 8 × 8 DCT transform coefficient block obtained through the discrete cosine transform is divided into a plurality of subblocks of a predetermined size, and a pattern is obtained by using a plurality of predetermined codewords provided in the memory for each of the divided subblocks. A method of updating a codeword in a system that outputs index values corresponding to input patterns of each subblock by performing vector encoding, the memory may include a first memory area for storing the plurality of preset codewords. And a second memory area for storing new codewords to be updated, and when the input pattern of the subblock is applied, the codewords of the reference pattern are searched from the first memory area to obtain an error value for the input pattern. Determining the least optimal reference pattern; Calculating a squared absolute difference value between the corresponding input pattern and the reference pattern and checking whether the calculated absolute difference value exceeds a predetermined threshold value; If it is determined that the calculated absolute difference value exceeds the preset threshold value, the new codeword corresponding to the input pattern is stored in a temporary buffer and the corresponding count value is increased by one. step; If the count value for the new codeword is checked and the value is not a predetermined count value, the process is repeated. If the checked count value and the preset count value match, the count value stored in the temporary buffer is stored. Reading a new codeword and storing it as an updated new codeword in the second memory area; And multiplexing the updated new codeword with index values corresponding to the input patterns of the subblocks, and outputting the multiplexed codewords. The method of claim 13, wherein the plurality of divided subblocks are four subblocks obtained by dividing each of the 8x8 DCT transform coefficient blocks into 4x4 units. . The method of claim 13, wherein the plurality of divided subblocks are 16 subblocks obtained by dividing each of the 8x8 DCT transform coefficient blocks into 2x2 units. . 16. The system of claim 13, 14 or 15, wherein the second memory area is a first-in first-out buffer memory that sequentially erases the updated new codewords according to an input order. How to update your codeword.