KR0159974B1 - Discrete cosine transform method and apparatus for image encoder - Google Patents

Abstract

The discrete cosine transforming apparatus of the video encoding system of the present invention is to enhance coding efficiency by adaptively applying a discrete cosine transform in a frame or in a field according to the characteristics (movement amount) of an input image in a video code and a system. To this end, the present invention detects a moving amount of an image in each block that has been converted to a predetermined size with respect to a video signal to be encoded, and performs on-field discrete cosine transform and in-frame discrete cosine transform based on the result of the moving amount detection. By selectively applying, the coding efficiency can be increased not only for still images but also for moving images, and for real-time design in application is easy.

Description

Discrete Cosine Transform Method and Apparatus for Video Coding System

The present invention relates to a discrete cosine transform (DCT) device employed in a video encoding system. In particular, the present invention relates to a video encoding in which a discrete cosine transform mode signal can be adaptively determined according to characteristics of an input image. A method and apparatus for discrete cosine conversion of a system.

In general, when a video signal and an audio signal, such as a video telephone, an HDTV, or a digital VCR, are to be processed as a digital signal, the digital data transmitted by using a high rate data compression method is compressed to reduce the enormous amount of data. That is, it is encoded and transmitted. At this time, the video signal and the audio signal are respectively encoded by different encoding devices due to the characteristics of the signals, and in this encoding, the encoding of the video signal, which generates more data than the audio signal, is particularly important. can see. Representative coding techniques for compressing video signals include, as is well known in the art, Motion Compensated Prediction (hereinafter referred to as MCP) and transform coding methods.

In other words, it can be said that the most important thing in the encoding of the video signal in such compression encoding is to reduce the amount of data to be transmitted. Therefore, when transmitting a video signal, the transmitting side encodes the image signal using spatial and temporal correlation lines of the video signal, and then transmits the image signal to the receiving side through a transmission channel.

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 (hereinafter, abbreviated as DCT), and performs motion estimation, prediction, and the like. By eliminating temporal redundancy of the video signal using differential coding, the video signal can be efficiently compressed.

These methods are now widely used in Moving Picture Experts Group (MPEG), H.261 (TV conferencing / telephone coding standard), and fully digital HDTV systems proposed by the US Federal Communication Commission (FCC). These mostly consist of MCP, CDT, quantization, entropy coding, and so on.

On the other hand, in successive images, there are typically three types of redundancy, that is, statistical redundancy in addition to the time redundancy and spatial redundancy as described above. Therefore, the compression of the video signal can be achieved by removing them. That is, as an example, in motion compensation prediction / transform coding, the MCP is for removing in-frame time redundancy present in a continuous video signal such as a TV.

However, the MCP error signal obtained through the MCP has almost eliminated the redundancy in the frame, but the spatial redundancy remains, so a conversion technique is used to remove the spatial redundancy of the MCP error signal. As a conversion method, a DCT having a high energy concentration efficiency is widely used for a signal having a high spatial correlation. The transformed DCT coefficients are quantized using different quantization intervals for each coefficient in consideration of the characteristics of the human visual system.

In general, quantization intervals are increased in coefficients in the high frequency region, and quantization intervals are reduced in coefficients in the low frequency region. In this quantization process, the amount of data is substantially compressed, and the amount of distortion introduced in proportion to the degree of compression is relatively increased. Although the temporal and spatial redundancy in the DCT coefficients of the quantized image data obtained through this process is removed, the statistical redundancy existing in each coefficient remains.

Therefore, in most coding systems, entropy encoding is performed using statistical characteristics of DCT coefficients to remove statistical correlations present in each coefficient. Distortion is not introduced in this process, and typical methods include Huffman coding and arithmetic coding, UVLC (Universal Variable Length Coding), and MUVLC (Modified UVLC).

In addition, in order to reduce the bit rate required to represent the quantized DCT coefficients, a method of entropy encoding a symbol (r, 1) in which a zero-length r and a nonzero coefficient 1 are combined while zigzag-scanning a DCT section is widely used. have.

FIG. 1 is a schematic block diagram of a conventional discrete cosine converting apparatus employed in a typical conventional video encoding system, and includes a video format converting unit 1 and a discrete cosine converting unit 2. As shown in FIG. That is, according to the conventional discrete cosine converting apparatus, the input image is converted into an 8 × 8 or 16 × 16 size block by the image format converting unit 1 and then applied to the discrete cosine converting unit 2 to convert the discrete cosine. This operation eliminates the redundancy of the data existing in the space area.

That is, the image format converter 1 included in the conventional discrete cosine converting apparatus 1 forms blocks in units of frames. For the 16 × 16 block, the video format for discrete cosine conversion as shown in FIG. Similarly, {p, (1, j), p (3, j),... , p (15, j); 1≤j≤16} are pixels corresponding to the first field, and {p (2, j), p (4, j),... , p (16, j); 1≤j≤16} are pixels corresponding to the second field. Accordingly, the image format converter 1 applies a discrete cosine transform for encoding to four 8x8 blocks configured in this manner.

However, as described above, when the discrete cosine transform is applied, there is a large difference in the movement of an object in a field when encoding a video such as a TV, so that a difference between pixels occurs between adjacent lines in a block, and thus the correlation between pixels decreases. When DCT is applied in a frame to an actual image, there is a problem in that coding efficiency is actually lowered.

Accordingly, the present invention solves the problems of the prior art described above, and in the video encoding system, the coding efficiency can be improved by adaptively applying the discrete cosine transform in the frame or the field according to the characteristics of the input image. Its purpose is to provide a discrete cosine conversion method.

Another object of the present invention is to provide a discrete cosine transforming apparatus of a video encoding system that can selectively apply a discrete cosine transform of an input image to be encoded in a field or a frame based on the amount of movement of the corresponding image.

According to an aspect of the present invention, an input image frame to be encoded is divided into a plurality of predetermined N × N frame blocks, and each pixel in the block is divided among the divided N × N frame blocks. In the discrete cosine transform method of a video system for compression coding in consideration of correlation, an input N × N frame block to be currently encoded is divided into even field and odd field blocks of each M × N, and the separated M × N even A first step of detecting movement amount information for the input N × N frame block by comparing a pixel average value difference between a field and an odd field block and a dispersion value calculated by the input N × N frame block; Generating a field-based transform encoding mode signal or a frame-based transform encoding mode signal for adaptive transform encoding of the input N × N frame block based on the detected movement amount information; When the field-based transform encoding mode signal is generated, the input N × N frame block is divided into even field blocks and odd field blocks of each M × N, and transform encoding is performed on each of the separated even field and odd field block data. Generating a second image data group that is transform-coded on a field basis by performing a method; Generating a second image data group that is transform-coded on a frame basis by performing transform encoding on the input N × N frame block data when the frame-based transform encoding mode signal is generated; And a fifth process of adaptively multiplexing and transmitting the generated field-based transform encoding mode signal or frame-based transform encoding mode signal and the encoded first image data group or encoded second image data group. Provides a discrete cosine conversion method.

According to another aspect of the present invention, an input image frame to be encoded is divided into a plurality of predetermined N × N frame blocks, and for each of the divided N × N frame blocks, each pixel in the block A discrete cosine transform apparatus of a video system for compression encoding in consideration of correlation, comprising: image format conversion means for dividing a current video frame input for encoding into a plurality of predetermined N × N frame blocks; The divided current N × N frame blocks are respectively divided into M × N even field and odd field blocks, and the pixel average value difference between the separated M × N even field and odd field blocks is calculated from the input N × N frame block. Movement amount detecting means for generating a field-based transformed encoding mode signal or a frame-based transformed encoding mode signal for adaptive transform encoding of the current N × N frame block based on the amount of movement obtained through comparison between one dispersion value; When the field-based transform encoding mode signal is generated, the current N × N frame block is divided into even field blocks and odd field blocks of each M × N, and transform encoding is performed on each of the separated even field and odd field block data. First discrete cosine transforming means for generating field-based transform-coded image data by performing a method; And second discrete cosine transform means for performing transform encoding on the current N × N frame block data to generate frame-based transform-coded image data when the frame-based transform encoding mode signal is generated. A cosine converter is provided.

1 is a schematic block diagram of a conventional discrete cosine transform apparatus employed in a typical typical video encoding system.

2 is a video format diagram for discrete cosine conversion.

3 is a block diagram of a discrete cosine transform apparatus of a video encoding system according to an exemplary embodiment of the present invention.

4 is a detailed block diagram of the motion detector shown in FIG.

* Explanation of symbols for main parts of the drawings

31: Image format converter 32: Movement detector

33,36: switching part 34,35: discrete cosine conversion part

41: Field Separator 42,43: Average Calculator

44,47: Subtractor 45: Absolute Value Calculator

46: Variance Calculator 48: Code Recognizer

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 core technical idea of the present invention is a frame-based discrete cosine transform and a first transform path (the path of the first discrete cosine transform unit 34 in FIG. 3) that performs field-based discrete cosine transform on an input image. Two discrete cosine transform paths of a second transform path (path on the side of the second discrete cosine transform unit 35 in FIG. 3) to perform the first transform path or the second transform path based on the detected amount of movement in the input image. 2 The transform path selects the output adaptively, i.e., the output of the first transform path (field-based transform-coded output) is selected as the final coded output if the input image has a large amount of movement, and the second if the input image has a small amount of movement. Output of the transform path (selected as frame-based transform encoding output).

At this time, the movement amount detection of the input image is determined based on the difference signal between the intra-frame dispersion values calculated using the absolute value of the average value difference between the two fields in which the frames are separated and the standard deviation indicating the average of the pixels in the frame and the conversion rate between the pixels. The present invention determines whether the amount of movement of the input image is large or small according to whether the sign of the difference signal between the absolute value of the mean value difference and the variance value, that is, the sign value of the difference signal is + or-, In response to the switching control signal determined by the determination result, field-based change coding (first transform path) or frame-based transform coding (second transform path) is adaptively selected.

Here, the difference signal between the absolute value and the variance value of the mean value difference means that the input image is a moving image having a certain amount of movement, and the difference signal between the absolute value and the variance value of the mean value difference is-value. Having a means that the corresponding moving image is an image in which the moving amount is close to the still image of JR.

That is, the present invention has a better coding efficiency than applying the discrete cosine transform in a frame in the case of a still image, and applies the discrete cosine transform in the field in the case of a video. It is possible to improve the coding efficiency in the discrete cosine transform, which is the object of the present invention, by performing adaptive coding in consideration of the coding characteristics of an image having a better coding efficiency than applying the discrete cosine transform in a frame. have. In other words, the present invention selectively applies field-based transform coding or frame-based transform coding according to the nature of the input video (i.e., the amount of movement), so that video signals having various characteristics distributed from still images to video are mixed. High coding efficiency can be achieved.

3 is a block diagram of a discrete cosine transform apparatus of a video encoding system according to the present invention.

As shown in the figure, the discrete cosine converting apparatus of the present invention includes an image format converter 31, a motion detector 32, first and second switching units 33 and 36, and first and second discrete cosine converters. (34,35).

Referring to FIG. 3, the image format converter 31 performs substantially the same function as the image format converter 1 of the above-described conventional apparatus, and for discrete cosine transform (DCT) at a later stage. The input image is divided into N × N block sizes, for example, 8 × 8 or 16 × 16 blocks, and each of the converted blocks is provided as an input of the movement detector 32 and the first switching unit 33, respectively. do.

Also, the movement detector 32 separates the absolute value of the average value difference between the two fields obtained by dividing a frame of the input image into two fields (that is, an even field and an odd feed), and a standard deviation indicating the average of the pixels in the frame and the rate of change between the pixels. On the first transform path (first discrete cosine transform unit 34 side path) or on the second transform path (second discrete cosine transform unit 35 side) based on the sign of the difference signal between the intra-frame dispersion values calculated using A switching control signal for selecting a path) as an encoding path of the input image, and providing the switching control signal to the first switching unit 33 and the second switching unit 36, respectively. Accordingly, a detailed operation process of detecting the movement amount of the input image will be described in detail with reference to FIG. 4 showing the detailed block configuration.

FIG. 4 is a detailed block diagram of the movement detector 32 shown in FIG. 3, which includes a field separator 41, two average value calculators 42 and 43, two subtractors 44 and 47, and an absolute value. A calculator 45, a variance calculator 46, and a sign recognizer 48 are included.

Referring to FIG. 4, the field separator 41 splits an input image divided into a plurality of N × N frame blocks provided from the image format converter 31 shown in FIG. 3 into an even field and an odd field. Here, the even field of the separated M × N field block is provided to the first average value adder 42, and the odd field of the separated M × N field block is provided to the second average value adder 43. Here, assuming that the frame block is a 16x16 block, each field block will be an 8x16 block.

Further, the first average value calculator 42 calculates an average value for even field blocks of M × N, and the second average value calculator 43 calculates an average value for odd field blocks of M × N. The average value of the even field block and the average value of the odd field block calculated at are provided to one side and the other side input of the first subtractor 44, respectively.

Next, the first subtractor 44 calculates the difference value, that is, the average value difference between the two field blocks by subtracting the average value of the even field block and the average value of the odd field block, and provides the difference value to the absolute value calculator 45 of the next stage. In the absolute value calculator 45, the absolute value of the calculated average value difference is taken and provided as one input of the second subtractor 47. That is, one side input of the second subtractor 47 is provided with an absolute value of a difference between two average values calculated in each field block obtained by dividing the input frame block into two field blocks.

On the other hand, in the dispersion calculator 46, the conversion rate between the average of the pixels and the pixels for each frame block of the input image frame divided into a plurality of N × N blocks provided from the image format converter 31 shown in FIG. The variance value according to the movement amount of each frame block is calculated by using the standard deviation, and the variance value of each N × N frame block calculated here is provided to the other input of the second subtractor 47.

Accordingly, the second subtractor 47 calculates the output difference value by subtracting between the absolute value of the mean value difference provided from the absolute value calculator 45 described above and the variance value provided from the variance calculator 46, wherein The calculated output difference value is a difference value having a + sign or a-sign according to the magnitude of the absolute value or the variance of the mean value difference, that is, if the absolute value of the mean value difference is relatively larger than the variance value, the output difference will be a difference value having a + sign. If the variance value is relatively greater than the absolute value of the mean value difference, the output difference will be a difference value with a minus sign.

Here, the output difference having a + value means that the input image to be encoded is a moving image having a certain amount of movement, and the output difference having a-value means that the input image currently to be encoded is closer to a still image with less movement amount. It means that the image is close.

On the other hand, considering the coding efficiency according to the characteristics of the conventional transform coding (DCT) for the image, applying the transform coding on a frame basis in the case of a small moving image is applied to the field-based transform coding in terms of coding efficiency It is already well known in the art that the field-based transform coding is more advantageous than the frame-based transform coding in terms of encoding efficiency in the case of a large moving image.

Therefore, in the code recognizer 48, when the sign of the output difference provided from the second subtractor 47 is + in consideration of the characteristics of the above-described transform coding, a field-based switching control signal corresponding thereto, for example, To generate a high level switching control signal and simultaneously provide it to the first and second switching units 33 and 36 of FIG. 3, on the contrary, if the sign of the output difference provided from the second subtractor 47 is -value, A frame-based switching control signal, for example, a low level switching control signal, is generated and simultaneously provided to the first and second switching units 33 and 36 of FIG. 3. That is, the code recognizer 48 generates an actual mode switching control signal for determining whether to perform field based transform encoding mode or frame based transform encoding on the input image block.

Although not illustrated in FIG. 4, the code recognizer 48 generates a mode signal for indicating whether the encoding mode of the corresponding input image block is a field based transform encoding mode or a frame based transform encoding mode, and thus not shown. The mode signal is multiplexed together with the image data transformed and encoded in units of blocks that are subject to transform encoding mode switching, and then transmitted to the transmission channel.

At this time, the first and second switching units 33 and 36 shown in FIG. 3 connect the contacts ab and a'-b 'when the high level switching control signal is provided from the code recognizer 48. Field-based transform coding mode, and connects the contacts ac and a'-c '(i.e., frame-based transform coding mode) when a low level switching control signal is provided from the code recognizer 48.

That is, the first switching unit 33 switches the input image, i.e., the output of the N × N frame blocks, provided from the image format conversion unit 31. When the input image block is a signal having a large amount of movement, The first discrete cosine converting unit 34 converts the input image (i.e., N × N frame) by connecting the contacts ab based on the field switching control signal (i.e., the high level switching control signal) provided from the code recognizer 48. The input image block by connecting the contact ac based on a frame switching control signal (i.e., a low level switching control signal) provided from the code recognizer 48 of FIG. 4 when the input image block is a signal having a small amount of movement. (I.e., N × N frame blocks) are provided to the second discrete cosine transforming section 35.

Meanwhile, in the field-based transform coding mode, the first discrete cosine transforming unit 34 may convert two N × N frame blocks provided through the first switching unit 33 into two M × N field blocks (even field). Blocks and odd field blocks), i.e. two frame blocks of 8x16 when the frame block is a 16x16 block, and the discrete two field block data are well known in the art for each of these separated fieldblock data. By performing the cosine transform (DCT), coded field-based DCT transform coefficients (that is, transform-coded image data) are generated and provided to the contact point b 'of the second switching unit 36.

Therefore, since the second switching unit 36 maintains the connection state of the contacts a'-b 'in the field-based transform encoding mode, the field-based DCT transform coefficients encoded by the first discrete cosine transform unit 34 described above. These are finally output as transform-coded image data of the input image block and provided to a transmitter not shown.

On the other hand, in the second discrete cosine transform unit 35, in the frame-based transform encoding mode, discrete N-N frame block data provided through the first switching unit 33 described above is well known in the art. By performing the cosine transform (DCT), encoded frame-based DCT transform coefficients (that is, transform-coded image data) are generated and provided to the contact point c 'of the second switching unit 36.

Accordingly, since the second switching unit 36 maintains the connection state of the contacts a'-c 'in the frame-based transform encoding mode, the frame-based DCT transform coefficients encoded by the second discrete cosine transform unit 35 are described. These are finally output as transform-coded image data of the input image block and provided to a transmitter not shown.

In other words, the discrete cosine transform scheme of the present invention is adapted to apply field-based transform coding to an image having a large amount of movement and to apply frame-based transform encoding to a small image with a large amount of movement based on a result of detecting the amount of movement of an input image. By employing a conventional transform encoding technique, such a discrete cosine transform technique of the present invention can be applied, for example, to a digital VCR, a video telephone, or the like by encoding techniques such as JPEG and MPEG.

As described above, according to the present invention, by selectively applying the field-based discrete cosine transform or the frame-based discrete cosine transform based on the characteristics of the input image, that is, the amount of movement of the image, in consideration of encoding efficiency in conventional transform encoding, It is possible to realize high coding efficiency for a video signal having various characteristics distributed from a still picture to a video, and to easily implement a real-time design in an application.

Claims (9)

Discrete cosine conversion method of a video system for dividing an input video frame to be encoded into a plurality of predetermined N × N frame blocks, and performing compression encoding on each of the divided N × N frame blocks in consideration of the correlation between pixels in the block. In the present invention, an input N × N frame block to be currently encoded is divided into even fields and odd field blocks of each M × N, and the pixel average value difference between the separated M × N even field and odd field blocks and the input N × A first step of detecting movement amount information for the input N × N frame block by comparing the variance value calculated in N frame blocks; Generating a field-based transform encoding mode signal or a frame-based transform encoding mode signal for adaptive transform encoding of the input N × N frame block based on the detected movement amount information; When the field-based transform encoding mode signal is generated, the input N × N frame block is divided into even field blocks and odd field blocks of each M × N, and transform encoding is performed on each of the separated even field and odd field block data. Generating a second image data group that is transform-coded on a field basis by performing a method; Generating a second image data group that is transform-coded on a frame basis by performing transform encoding on the input N × N frame block data when the frame-based transform encoding mode signal is generated; And a fifth process of adaptively multiplexing and transmitting the generated field-based transform encoding mode signal or frame-based transform encoding mode signal and the encoded first image data group or encoded second image data group. Discrete cosine conversion method. The method of claim 1, wherein the first step comprises: an eleventh step of calculating a first pixel average value in the separated M × N even field block and a second pixel average value in the separated M × N odd field block, respectively; A twelve step of generating a pixel average value difference by subtracting the calculated first and second pixel average values, and generating an absolute value of the generated pixel average difference; Calculating a variance value of the input N × N frame block using the calculated average and standard deviation, respectively, and calculating a mean and a standard deviation between the pixels in the input N × N frame block; And a fourteenth step of calculating a difference signal between the absolute value of the generated average difference and the calculated dispersion value, and detecting movement amount information for the input N × N frame block based on the calculated difference signal. Discrete cosine transform method of a video encoding system, characterized in that. The method of claim 1 or 2, wherein the input NxN frame block is an 8x8 block. 3. The method of claim 1 or 2, wherein the input NxN frame block is a 16x16 block. Discrete cosine conversion device of video system for dividing input video frame to be encoded into a plurality of predetermined N × N frame blocks, and compressing and coding each divided N × N frame block in consideration of the correlation between each pixel in the block. A video format conversion means for dividing a current video frame input for encoding into a plurality of predetermined N × N frame blocks; The divided current N × N frame block is divided into even field and odd field blocks of each M × N, and the pixel average value difference between the separated M × N even field and odd field blocks is calculated from the input N × N frame block. Movement amount detecting means for generating a field-based transformed encoding mode signal or a frame-based transformed encoding mode signal for adaptive transform encoding of the current N × N frame block based on the movement amount obtained through comparison between one dispersion value; When the field-based transform encoding mode signal is generated, the current N × N frame block is divided into even field blocks and odd field blocks of each M × N, and transform encoding is performed on each of the separated even field and odd field block data. First discrete cosine transform means for generating field-based transform-coded image data by performing a plurality of operations; And second discrete cosine transform means for performing transform encoding on the current N × N frame block data to generate frame-based transform-coded image data when the frame-based transform encoding mode signal is generated. Cosine converter. 6. The method of claim 5, wherein the first discrete cosine converting means converts the current N × N frame block data into discrete cosine based on a field when the moving amount of the image in the current N × N frame block is large, and the second discrete cosine. And the converting means performs discrete cosine transform of the current N × N frame block data on a frame basis when the amount of movement of the image in the current N × N frame block is small. 7. The discrete cosine transform apparatus according to claim 5 or 6, wherein the current NxN frame block is an 8x8 block. 7. The discrete cosine transform apparatus according to claim 5 or 6, wherein the current NxN frame block is a 16x16 block. 7. The apparatus according to claim 5 or 6, wherein the movement amount detecting means comprises: field separating means for separating the current input N × N frame block into an even field and an odd field block of each M × N; First and second average value calculating means for calculating a first pixel average value in the separated M × N even field block and a second pixel average value in the separated M × N odd field block, respectively; Absolute value calculating means for generating a pixel average value difference by subtracting the calculated first and second pixel average values, and calculating an absolute value of the generated pixel average difference; Dispersion value calculating means for calculating an average of the pixels in the current N × N frame block and a standard deviation between the pixels, and calculating a dispersion value of the current N × N frame block using the calculated average and standard deviation; Calculates a difference signal between the calculated absolute value of the average difference and the calculated dispersion value, detects the movement amount with respect to the input N × N frame block based on the calculated difference signal, and based on the detected movement amount And a mode signal generating means for generating the field-based transform coded mode signal or the frame-based transform coded mode signal.