KR20000045076A - Compression coder and decoder of video signal and method thereof - Google Patents

Abstract

PURPOSE: A compression coder and a decoder of a video signal and a method thereof are provided to compress digital input images in image units and have compatibility with MPEG application products. CONSTITUTION: A compression coder and a decoder of a video signal and a method for the same comprise a discrete cosine transform unit(104), an initial quantizer(108), a code capacity controller(112) and a real quantizer(122). The discrete cosine transform unit performs a discrete cosine transform for input signals in order to provide a discrete cosine transform coefficient. The initial quantizer quantizes the discrete cosine transform coefficient by using a quantization matrix in order to provide an initial quantization data. The code capacity controller quantizes in parallel the initial quantization data by a quantization step size and outputs a code capacity near to a target code capacity of code capacities. The real quantizer provides compressed data by quantizing the initial quantization data.

Description

Compression coder and / or decoder of video signal and method thereof

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the field of compression encoding and / or decoding of video signals, and more particularly, to an encoder and / or decoder having a high performance compression encoding scheme in a digital recording / reproducing apparatus.

After the completion of the Moving Picture Experts Group (MPEG-1) standardization, the MPEG-2 standard was born, which required a standard for realizing high image quality at higher bit rates. This MPEG-2 standard is considered for application not only to storage media but also to broadcast media, and currently targets higher quality than television (TV) quality, and can be extended to high definition television (HDTV) quality. Therefore, it has a wide range of applications in various fields such as communication, home appliances, computers, and broadcasting. As the importance of MPEG, which is a standard for video compression, has been increasing day by day, the application products using it have also become very diverse. However, such a compression coding scheme is difficult to apply to an application product that needs to fix the amount of data compressed in one or several screen units. Therefore, in order to maximize the performance of the compression encoder of such an application product, a compression encoder that is as close as possible to the target data amount is essential.

In accordance with the trend of demanding high quality, an independent standard was developed by Japan's leading compression coding method of digital video camcorder (DVC), which records and plays digital video from cameras on videotape. As a result, the recorded content cannot be viewed on a digital television or a multimedia system adopting the MPEG encoding method that will be spread in the future. However, after the world standards of next-generation digital video cassette recorders (VCRs) and camcorders have been decided, SD (standard definition) class camcorders are rapidly spreading among Japanese companies. Since the picture quality is made in units of screens, the quality of the picture depends on the amount of compressed code. will be.

On the other hand, the HD Digital VCR Council was established in 1993 to propose the basic specifications of the digital video format and to discuss the HDTV specifications based on these basic specifications. Was enacted. The main points of this standard are the rotating head method (helical scan), the compression method based on the Digital Cosine Transform (DCT), the 1/4 inch wide video tape, etc., and the HD baseband format is the standard for HDTV. The sampling frequency is 50.4 MHz, which is three times the SD class sampling frequency (13.5 MHz), and the video data recording rate is 50 Mbps, which is twice the SD class 25 Mbps. Currently, SD class digital video camcorders are commercially available.

In 1994, the DTV Working Group (WG) was launched in the United States to discuss the format for recording DTV signals, and in Europe, the DVB WG was launched to discuss the format for recording DVB (Digital Video Broadcasting) signals. .

When recording a DTV signal compressed at a data rate of 19.3 Mbps to a video tape in SD format, a 25 Mbps video data rate in SD is recorded. In addition, the data rate of the DVB signal is controlled to 10 Mbps or less, and when the DVB signal is recorded on an SD format video tape, it is recorded at 12.5 Mbps, which is half of the SD data rate, or 6.25 Mbps, which is 1/4. Since DTV adopts MPEG-2, it is necessary to use a digital recording and reproducing apparatus using a format for recording data compressed by the MPEG-2 method as it is. Here, the DTV signal is composed of data compressed in MPEG-2 of 19.3 Mbps and compressed data to be repeatedly recorded in a specific area for high-speed search and recorded on a video tape. The high-speed search data to be repeatedly recorded uses a margin equal to the difference between the recording rate and the compression rate (25 Mbps to 19.3 Mbps) for the high-speed search.

Apart from digital VCRs that record these DTV signals, the current DVC compresses and encodes to its own format without any compatibility with MPEG-1 and MPEG-2, and plays a major role such as high-speed search and screen unit editing. This is possible because it has a compression coding scheme of a unit of picture and an independent encoded data structure of fixed size called a segment. That is, it is possible to use a compression coding technique that enables high-speed search and editing of screen units while maintaining high quality. However, as the use of MPEG, the video compression standard, increases, its importance and application are increasing, so the use of this standard in DVC will have relatively more consumers. In other words, the digital video camcorder will be further utilized if the DVC which is compatible with the DTV or multimedia application products adopting the MPEG-2 compression method and the DVC, which is the main function of the DVC, enables high-speed navigation and picture-by-screen editing.

In this case, it is difficult to control the code amount of a screen unit only by the MPEG encoding method, and thus a method for compensating for this is necessary. Because MPEG-2 compresses by using the correlation between screens in GOP (group of picture) unit consisting of several screens, if the compressed data of such structure is simply recorded on tape as it is, variable speed playback The composition of the screen is difficult and does not play properly during (High Speed Search).

In summary, when recording a DTV signal composed of MPEG-2 compressed data and compressed data to be repeatedly recorded on videotape, one of the main functions of a digital VCR is fast navigation, but a single screen called a GOP composed of multiple screens. Since compression coding is performed in an independent unit, editing of the screen unit is impossible and the reproduction quality is lower than that of the SD class.

In addition, the existing SD class digital video camcorder compresses by screen unit and compresses independently by segment unit consisting of a predetermined number (typically 5 macroblocks), and enables screen unit and high speed search. There was a problem that is not compatible with the device.

In order to solve the above problems, it is an object of the present invention to provide an encoder and / or decoder that compresses a digital input video signal on a screen basis and is compatible with MPEG application products.

Another object of the present invention is to provide a compression encoding and / or decoding method of a video signal for a digital recording / reproducing apparatus such as a digital camcorder or a digital VCR, which requires image processing in units of screens.

In order to achieve the above objects, the encoder according to the present invention performs discrete conversion of an input video signal to provide discrete transform coefficients, and quantizes discrete transform coefficients using an initializing quantization matrix to provide initial quantized data. An initial quantizer and a code that provides a quantization step size that outputs the result closest to the target code amount among the respective code amounts obtained by quantizing the initial quantized data in parallel using a plurality of quantization step sizes, as a real quantization step size. And a real quantizer for quantizing the initial quantized data by the amount controller and the actual quantization step size to provide compressed data.

In addition, the coder of the present invention performs variable length coding by temporarily storing the initial quantized data while providing the actual quantized step size in the code amount controller, and performing continuous length and variable length coding on the compressed data. The apparatus may further include a continuous and variable length encoder for providing data, and a basestream encoder for generating a basestream by adding an additional header and outputting the basestream as a bitstream.

The decoder according to the present invention extracts the elementary stream from the transmitted bitstream, removes the additional header from the elementary stream, and provides a video stream, and continuously decoded data by variable-length and continuous-length decoding the video stream. Variable- and continuous-length decoders to provide the inverse quantizer and inverse quantized data to inversely quantize the inverse quantized data using the quantization step size used for encoding the inverse quantized data. And an inverse discrete conversion unit for outputting data.

In addition, the encoding method according to the present invention comprises the steps of providing discrete transform coefficients by discretely transforming an input video signal, quantizing the discrete transform coefficients using an initializing quantization matrix to provide initial quantized data, initial quantized data Determining the quantization step size that outputs the result closest to the target code amount from each code amount obtained by quantizing in parallel using a plurality of quantization step sizes as the actual quantization step size and the initial by the determined actual quantization step size Quantizing the quantized data to provide compressed data.

In addition, the encoding method of the present invention provides continuous-length and variable-length encoding of compressed data to provide variable-length encoded data, and generates a basestream by adding an additional header to output the variable-length encoded data as a bitstream. It further comprises a step.

The decoding method according to the present invention extracts an elementary stream from a transmitted bitstream, removes an additional header from the elementary stream, and provides a video stream, and provides variable-length decoded data by variable-length and continuous-length decoding of the video stream. And inversely quantizing the continuous-length decoded data by the quantization step size used for encoding to provide inverse quantized data, and inversely discretely transforming the inverse quantized data and outputting the original image data. It is done.

1 is a block diagram according to an embodiment of an encoder according to the present invention.

2 illustrates an example of an input image format.

3 illustrates another example of an input image format.

4 illustrates another example of an input image format.

FIG. 5 is a detailed block diagram of the code amount controller shown in FIG. 1.

FIG. 6 is a view for explaining the concept of I-scan and D-scan performed in the I / D scan unit shown in FIG. 1.

7 (a) and 7 (b) are diagrams for explaining an example of an I-scan and a D-scan order of the I / D scan unit shown in FIG.

8 is a block diagram according to an embodiment of a decoder according to the present invention.

Hereinafter, with reference to the accompanying drawings will be described a preferred embodiment of the compression encoder and / or decoder and the method of the video signal according to the present invention.

First, in order to assist in understanding the present invention, a video compression encoding and decoding process for MPEG-2 is defined in ISO / IEC 13818-2. This ISO / IEC 13818-2 describes the meanings of terms used to describe the decoding process, the grammar and hierarchical structure of bit strings, the meaning of various parameters, variable length decoding (VLD), inverse scan (ISCAN), inverse quantization (IQ), A description of processes such as Inverse Discrete Transform (IDCT) and Motion Compensation (MC) is included.

In addition, in order to increase the compression rate by eliminating temporal redundancy with the previous frame, the MPEG-2 finds the macroblock (reference block) and the closest block (matching block) of the current frame in the previous frame, and then A motion prediction method of generating and encoding a prediction error that is a difference between a motion vector that is a position difference and a pixel value of a reference block and a matching block is used. Accordingly, the MPEG-2 quantizer stores macroblocks (intra macroblocks) that do not use motion prediction, macroblocks that use motion prediction (inter macroblocks), quantization intervals having different weights according to luminance signals and color signals. Quantization table is used.

However, since the present invention does not use a technique that uses the correlation between the current screen and the previous screen because of the editing of the screen unit, such a coding method for motion prediction and compensation is not used. That is, the screen-based editing function is enabled using only I (intra) pictures. The bitstream output from the encoder of the present invention can be output as a bitstream of the MPEG-2 system layer format, thereby maintaining compatibility with MPEG-2.

In FIG. 1, which is a block diagram according to an embodiment of the present invention, the preprocessor 102 subsamples a 4: 2: 0 format when an input video signal has a format of 4: 2: 2. Low-pass filtering is performed to perform subsampling on the color components. In other words, unlike the luminance signal that includes most of the information that can be understood by the human eye in the input image data, the color signal has low frequency components, so low-pass filtering and subsampling are performed before encoding. This is because the color component is less important than the luminance component in the compression process, and as a result, a high performance compressor can be realized by reducing the burden on the compression ratio of the compression process.

Here, the image format input to the preprocessor 102 may vary. An example of the input image format is illustrated in FIG. 2, and the size of the input image is 720 pixels. × 480 lines, one DCT block size is 8 pixels × 8 lines, the macro block (MB) is composed of four DCT blocks of the luminance component and two DCT blocks of the color component, and the slice (SMB) shows the form of the input data in the case of five macro blocks. , So 1 frame equals 270 (= 9 × 30). Unlike the general size, the slice structure is composed of a small number of macroblocks and compressed by the present invention for fast searching.

Another example of the format of the input image is shown in FIG. 3, and the size of the input image is 720 pixels. × 480 lines, the slice (SMB) shows the form of input data in the case of three macroblocks, so one frame is 450 (= 15 × 30).

Another example of the format of the input image is shown in FIG. 4, where the size of the input image is 640 pixels. × 480 lines, the slice (SMB) shows the form of input data in the case of five macroblocks, so one frame is 240 (= 8 × 30).

The DCT and ZZ (Zig Zag) scanning unit 104 performs a zigzag scan of the DCT coefficients obtained by performing DCT on the image data provided from the preprocessor 102. DCT, a transformation that concentrates energy in the low frequency region, is used not only for MPEG-1 and MPEG-2 but also for many other international video signal compression standards. Since the human eye is more sensitive to low frequency components than high frequency components, it eliminates many high frequency regions. However, the image quality deteriorated by a person uses a small visual characteristic. DCT converts spatial pixel values into the frequency domain. DCT itself cannot compress image data and is always used with quantization.

Since the present invention does not use motion estimation (ME), the quantizer only needs to have a quantization table for the luminance signal and the color signal. In addition, since the present invention does not use motion prediction, a motion predictor having a large amount of computation is unnecessary, and the amount of hardware is reduced since a frame memory and a separate decoder for motion prediction are not used.

On the other hand, due to the characteristics of the camcorder, the code amount is fixed in units of screens, and the quantization process of adjusting the code amount is the most important part in determining the image quality during playback. Among the various parameters of the encoder used in MPEG-2, the contents of the main variables used in the present invention are as follows. First, intra_dc_precision represents the number of coding bits of an intra block DC coefficient. In the case of DC precision, the precision of DC quantization is expressed by 8 bits in MPEG-1, but the precision of 8/9/10/11 bits can be selected in MPEG-2 and this can be changed for each picture. In order to simplify the hardware, the encoder of the present invention is fixed with a precision of 10 bits, but there is some difference in the amount of code between using 8 bits and 11 bits, but there is no significant difference in terms of image quality. It also uses bits.

The change of the quantization matrix corresponding to the DCT block is allowed only in a sequence unit in MPEG-1, but can be changed in units of pictures in MPEG-2 and can be downloaded. In the present invention, only the default quantization matrix used in MPEG-2 is used. If q_scale_type, which changes the quantization step size, has a value of "0", it takes a table in which the step size [2-62] is changed linearly, which is compatible with MPEG-1. On the other hand, if the value is "1", the step size changes nonlinearly [1-112], and the effect is great when the quantization step size is finely adjusted. Such a nonlinear step structure is used in a high bit rate encoder. In the present invention, q_scale_type is fixed by encoding "1".

Therefore, in the present invention, the quantization process, which is a process of increasing the compression rate by quantizing the DCT transform coefficients, is performed in two steps for hardware simplification. The first step is the quantization step using the quantization matrix used in MPEG-2, ie the quantizer 106 is stored in the predetermined initial quantization table 108 and the DCT coefficients applied from the DCT and ZZ scan units 104. Quantize uniformly throughout the screen using the default quantization matrix, and its output is stored in frame buffer 110 during processing for the bitrate controller 112. If the input image is complex, the coded bit stream exceeds the predetermined code amount, so that the coefficient of quantization step size is increased to reduce the compressed code amount, and in the opposite case, the coefficient of the quantization step size is decreased to increase the compressed code amount relatively. Adjustment is necessary.

On the other hand, the digital video signal input to the encoder is usually compressed because the amount of data is too large to transmit the code amount or to record on the recording medium. Like a digital camcorder, a compression encoding method using a compression unit of a screen unit for functions such as screen unit editing or high-speed search is very different from the compression technique used in MPEG. That is, the code amount control of the encoding method of fixing the code amount on a frame basis shows a lot of difference from the general code amount control.

In addition, when the compressed data is recorded on a video tape, the code amount compressed in units of frames is kept at a predetermined target value or less and recorded. In the case of NTSC, 10 tracks are allocated to one frame, and in the case of PAL, 12 tracks are allocated. Therefore, if the code amount prediction is deficient and the compressed code amount falls short of the target value or remains, the image quality is greatly affected. In other words, if the compressed code amount is insufficient to the target value, deterioration of the image quality occurs during playback. If the code amount is larger than the target value, compression is forcibly terminated and the last part of the screen is cut off.

Due to these characteristics, the code amount adjustment of the degree of application or application of a general bit rate control method that is adjusted according to the characteristics of the input image (for example, activity) and the amount of code of the output buffer is immediately applied. It shows the limitation in playing the higher picture quality. According to the experiment, it is shown that the best compression method in terms of subjective and objective is to output the maximum code amount that does not exceed the target code amount of one screen in the case of compression coding on a screen basis.

On the other hand, in the DVC standard, since the code amount is fixed in units of segments composed of five macro blocks mainly for the purpose of fast searching, the next segment is not affected by the code amount. In the compression method of the present invention, a slice unit is composed of a predetermined number of macroblocks smaller than a general size (for example, five) for fast search, which is an essential function of a camcorder, to replace the role of an SD class segment unit.

However, although the size of the SD class segment is constant, the compression scheme of the present invention is variable. If the size of the segment, which is an independently decodable unit, is constant, it can show more efficient performance during fast searching and error concealment, but it is inferior to the variable coding method in terms of efficiency in compression reproduction. In addition, in the case of SD-class DVC, since the quantization is performed in units of DCT blocks, its characteristics are slightly different from those of MPEG compression, which is performed in units of macroblocks. In other words, the subjective characteristics of SD-class DVC may appear more predominantly in the edge component of a complex mobile in which quantization must be controlled in units of DCT blocks.

Next, a description will be given based on a second quantization step of obtaining and quantizing the actual quantization step size by the code amount control of the present invention, which is most distinguished from the existing encoder. The second quantization step uses the reference quantization step size and the neighboring quantization step size to determine the quantization step size closest to the target code amount among the respective code amounts obtained by performing parallel quantization on the data quantized by the initial quantization step size. To generate the actual quantization step size.

That is, the code amount adjuster 112 is composed of a plurality of quantizers 114, a plurality of code length counters 116, and an adjuster 118, and a detailed block diagram is shown in FIG. In FIG. 5, as an example of the plurality of quantizers 114, the first to third quantizers 142, 144, and 146 may store data quantized by the initial quantization step size provided from the quantizer 106 than the reference quantization step size. Quantization is performed in parallel by a step size mquant + 1 having a large " 1 ", a reference quantization step size mquant and a step size mquant-1 having a " 1 " smaller than a reference quantization step size.

The first to third counters 162, 164, and 166 of the plurality of code length counters 116 count the quantized data provided from the first to third quantizers 142, 144, and 146, and use the amount of code when each quantization step size is used. A count value indicative of is provided to the regulator 118. Here, sum_P and the second counter 164 are the sum of the first code amount in the case of using the quantization step size mquant + 1 which is one step (here, "1") larger than the reference quantization step size provided from the first counter 162. Sum the second code amount in the case of using the reference quantization step size (mquant) provided from, and use the quantization step size (mquant-1) that is "1" smaller than the reference quantization step size provided from the third counter 166. The third code amount in the case is called sum_M.

At this time, the adjuster 118 adjusts the reference quantization step size to be used in the next input screen based on the first to third code amounts and the target code amounts provided from the first to third counters 162, 164, and 166. The feedback size is fed back into the plurality of quantizers 114. Through this process, even if a scene change occurs when the screen of the input image changes, a reference quantization step size suitable for the scene can be extracted immediately.

In addition, the adjuster 114 determines the actual quantization step size according to the first to third code amounts sum_P, sum, and sum_M calculated by the first to third counters 162, 164, and 166. It is divided into a case of performing quantization while performing an I / D scan by the I (Increasing) / D (Decreasing) scan unit 120 and a case of quantizing without performing an I / D scan.

That is, the code amount controller 112 is obtained by performing parallel quantization with a reference quantization step size (mquant), one step size larger than the reference quantization step size, and one smaller size (mquant + 1, mquant-1). Based on the code amount of one screen corresponding to each, it may be determined whether to scan the I / D as given in Equation 1 below. That is, when the target code amount TB is between one less than the reference quantization step size and the code amount due to a large amount (Bm> TB> Bp), the I / D scan unit 120 performs the I / D scan while performing the reference. The actual quantization is performed using the actual quantization step size whose quantization step size is adjusted according to the position of the screen.

In other cases, actual quantization is performed using the actual quantization step size obtained by adding or subtracting an offset to the reference quantization step size. That is, the code amount Bm when the target code amount TB is less than "1" than the reference quantization step size and the quantization step size that is "1" larger than the reference quantization step size are used. In the case of overflow or underflow not between the code amounts Bp in the case, it is necessary to change the reference quantization step size mquant by adding or subtracting the offset.

TB> Bm> Bp: change mquant to [mquant-offset_M]

Bm> TB> Bp: I / D scan

Bm> Bp> TB: change mquant to [mquant + offset_P]

where, Bm: sum_M, Bp: sum_P, TB: Target Bit Number

Equation 2 below shows an example of the actual quantization step size (re_mq) generated by adding an offset to the reference quantization step size due to overflow, and Equation 3 shows an offset to the reference quantization step size by underflow. An example of the actual quantization step size generated by subtracting is shown.

Tgap = B-TB, Pgap = B-Bp

if Tgap / 2> Pgap re_mq = i_mqaunt + 8

else if Tgap / 1> Pgap re_mq = i_mquant + 4

else re_mq = i_mquant + 1

where, Bm; Bit amount by mquant-1, B: Bit amount by mquant, Bp: Bit amount by mquant + 1, re_mq: real quantization stepsize

Tgap = TB-B, Mgap = Bm-B

if Tgap / 32> Mgap re_mq = i_mquant-8

else if Tgap / 16> Mgap re_mq = i_mquant-5

else if Tgap / 8> Mgap re_mq = i_mquant-3

else if Tgap / 4> Mgap re_mq = i_mquant-2

else re_mq = i_mquant-1

where, Bm; Bit amount by mquant-1, B: Bit amount by mquant, Bp: Bit amount by mquant + 1, re_mq: real quantization stepsize

Therefore, the quantization process of the second step is performed by the frame memory based on the new macroblock quantization step size (actual quantization step size) adjusted to the difference between the final code amount obtained by the code amount controller 112 and the target code amount. Quantize the data read from 110. That is, as described in Equation 1 above, the third code amount and the quantization step size in which "1" is larger than the reference quantization step size when the target code amount is less than "1" than the reference quantization step size are used. If it is between the first code amount in the case of use, the actual quantizer 122 may quantize while changing the determined reference quantization step size while performing I-scan or D-scan by the I / D scan unit 120. Otherwise, if overflow or underflow occurs, the frame buffer 110 in the actual quantizer 122 without passing through the I / D scan unit 120 by the actual quantization step size determined by Equation 2 or 3 below. Quantize the data read from Through this process, even if the scene of the input image changes, the quantization process having the result that the compressed code amount approaches the target code amount can be performed.

On the other hand, the I-scan or D-scan used in the I / D scan unit 120 compresses the data that is most closely compressed to the target code amount when the code amount is fixed and encoded by one screen or several screen units. This is a scan method used to re-encode a specific part such as the edge of the screen or the center part of the screen until the target code amount is output. I-scan means a scanning order in units of macroblocks for decreasing the amount of generated code, and D-scan means a scanning order in units of macroblocks for increasing the amount of generated code.

Therefore, the second quantization step is a quantization process that allows finer code amount control. Even when the first step of quantization is performed, the code amount often exceeds or falls short of the target value. Therefore, the quantization process is performed by varying the quantization step size in screen units or macroblock units in order to adjust the corresponding code amount in each case. Quantize by adjusting the actual quantization step size to output the amount.

On the other hand, since the transform coefficients that have undergone quantization with DCT include many coefficients having a value of "0" by quantization, the continuous field coder (RLC) 124 continuously stores the quantization coefficients provided from the actual quantizer 122. Encoding increases the compression rate. This continuous field coding outputs the number of consecutive "0" s and the value of the coefficient which has a non-zero value immediately following one symbol by using the characteristic of DCT transform coefficients with many "0" s.

The variable length coder (VLC) 126 uses the code table 128 to allocate the shortest codeword to the data that appears most probably among the continuous-length coded symbols, and to assign a long codeword to a symbol having a relatively low probability of occurrence. Allocate it to increase the compression rate in a stochastic way. The elementary stream (ES) encoder 130 temporarily stores the variable length coded data provided from the VLC 126 in the code buffer 132 and adds an additional header to the stored data to generate the elementary stream. To output in bitstream form.

FIG. 6 is a diagram for describing a concept of an I / D scan performed by the I / D scan unit 120 illustrated in FIG. 1. First, I-scan corresponds to a case where the code amount calculated by the code amount adjuster 112 is smaller than the target code amount that can be actually recorded, and is an encoding order in units of macroblocks. That is, if the compressed code amount by the determined reference quantization step size is smaller than the target code amount, the reference quantization step determined in macroblock units from the central part of the visually more important screen to fit the compressed code amount of the screen to the target code amount. The quantization process is performed again while changing the size, and the encoding process is performed until the target code amount is approached. At this time, the quantization step size to be re-encoded increases the amount of code output from the actual quantizer 122 by using one step lower than the determined reference quantization step size.

The D-scan is a coding order in macroblock units corresponding to the case where the code amount calculated by the code amount controller 112 is larger than the target code amount. That is, if the compressed code amount by the determined reference quantization step size is larger than the target code amount, the reference quantization step determined in macroblock units from the edge of the screen which is less visually important to fit the compressed code amount of the screen to the target code amount. The quantization process is performed again while changing the size, and the encoding process is performed until the target code amount is approached. In this case, the quantization step size to be re-encoded is one step higher than the determined reference quantization step to reduce the amount of code output from the actual quantizer 122.

7 (a) and 7 (b) show 44 rows in one screen. × FIG. 1 shows an example of a sequence of D-scan and I-scan when there are 30 (col) macroblocks. As shown in (a) of FIG. 7, the D-scan uses a step higher than the reference quantization step size in which the quantization step size is determined from the edge so that the actual code amount approaches the target code amount. As shown in FIG. 1, the I-scan actually quantizes the quantization step size from the center portion using one step lower than the determined reference quantization step size.

8 is a block diagram according to an embodiment of the decoder according to the present invention. The basestream decoder 202 extracts a basestream from a transmitted bitstream and removes an additional header added to the extracted basestream to remove the video. Provide a stream. The variable length decoder (VLD) 204 variable length decodes the video stream using the code table 206. The continuous field decoder (RLD) 208 continuously decodes the variable length decoded data, and the inverse quantizer (IQ) 210 uses the quantization step size used during encoding stored in the quantization table 212 in the continuous field decoded data. Multiply to give inverse quantized data.

The IZZ / IDCT unit 214 performs an inverse process of the zigzag scan on the inverse quantized data, inverse discrete transforms the inverse zigzag scanned data, and temporarily stores the original image data in the line buffer 216, and then selects a predetermined number of lines. The unit is provided to the post-processor 218. The postprocessor 218 performs the reverse process of the preprocessor 102 shown in FIG. 1 because the preprocessor 102 downsampled the 4: 2: 2 format signal to the 4: 2: 0 format. The post processor 218 outputs a video signal of the original 4: 2: 2 format by performing a corresponding upsampling.

The present invention divides the quantization process into a quantization matrix and a process of dividing the actual quantization step size into two, and first divides the input image signal into an initial quantization matrix once, and then performs the initial quantization in the part of quantizing in parallel or the part of actually performing the quantization process. The process of dividing into matrices can be omitted, which reduces the complexity of the hardware.

In addition, in the present invention, in order to perform the screen unit editing and the fast search function, the compressed code amount is fixed to a fixed size or less for each screen, and the unit capable of independent decoding for the fast search must be small in size. Therefore, the slice unit is composed of several macroblocks. In an embodiment of the present invention, the actual quantization step size is described in macroblock units, but may be a slice unit composed of several macroblocks.

The present invention can be effectively applied to digital cameras, camcorders or digital VCRs that require image processing on a screen basis by adopting a video signal compression coding scheme that can compress an input digital video signal into a predetermined size on a frame (screen) basis.

As described above, in the present invention, when applied to a digital VCR, the video processing can be performed on a screen basis by compression encoding and decoding on a screen (intra picture) basis. When applying to a DVC, the video signal is a standard of MPEG-2. There is an effect that is compatible with all devices using MPEG by compression encoding and decoding so as to be suitable for. In addition, the present invention improves the compression performance of the compression encoder, the image quality is improved and the hardware has a simple effect.

Claims (34)

In the method of compression encoding and / or decoding the input video signal: (a) discrete converting an input video signal to provide a discrete transform coefficient; (b) quantizing the discrete transform coefficients to provide initial quantized data using an initialization quantization matrix; (c) determining, as the actual quantization step size, a quantization step size that outputs a result closest to a target code amount among the respective code amounts obtained by quantizing the initial quantized data in parallel using a plurality of quantization step sizes; And (d) quantizing the initial quantized data by the determined actual quantization step size to provide compressed data. The method of claim 1, wherein step (c) (c1) quantizing the initial quantized data in parallel using a reference quantization step size, one step larger quantization step size, and one step smaller quantization step size to provide first to third quantized data; (c2) counting the first to third quantized data by a predetermined size unit to provide first to third code amounts; And (c3) determining an actual quantization step size using a quantization step size that outputs a result closest to a target code amount among the first to third code amounts. 3. The method of claim 2, wherein the step (c3) uses a first code amount obtained by using a quantization step size larger than the reference quantization step size and a quantization step size one step smaller than the reference quantization step size. The reference quantization step size is determined as the actual quantization step size if the second code amount obtained by the reference quantization step size is larger than the target code amount, and the second code amount is set to the target code amount. Varying the quantization step size of the edge of the screen that is less visually significant until it is close. The method of claim 3, wherein in step (c3), the quantization step size of the edge of the screen is changed to a value one step larger than the determined reference quantization step size. The method of claim 3, wherein step (d) (d1) scanning the initial quantized data from the edge of the screen in macroblock units; And (d2) quantizing the scanned initial quantized data using the determined reference quantization step size, wherein the edge of the screen uses a quantization step size larger than the determined reference quantization step size. 3. The method of claim 2, wherein the step (c3) uses a first code amount obtained by using a quantization step size larger than the reference quantization step size and a quantization step size one step smaller than the reference quantization step size. The reference quantization step size is determined as the actual quantization step size if the second code amount obtained by the reference quantization step size is smaller than the target code amount, and the second code amount is the target code. Varying the quantization step size from the center of the visually important screen until it is close to the quantity. 7. The method of claim 6, wherein in step (c3), the quantization step size of the center portion of the screen is changed to a value one step smaller than the determined reference quantization step size. The method of claim 6, wherein step (d) (d1) scanning the initial quantized data from the center of the screen in units of macroblocks; And (d2) quantizing the scanned initial quantized data using the determined reference quantization step size, wherein the quantization step size of the center portion of the screen uses a quantization step size smaller than the determined reference quantization step size. 3. The method of claim 2, wherein the step (c3) uses a first code amount obtained by using a quantization step size larger than the reference quantization step size and a quantization step size one step smaller than the reference quantization step size. If the second code amount obtained by the reference quantization step size is not larger than the target code amount, the offset is added to the reference quantization step size to provide the actual quantization step size. How to. 10. The method of claim 9, wherein the offset value is obtained by the following formula. Tgap = B-TB, Pgap = B-Bp if Tgap / 2> Pgap re_mq = i_mqaunt + 8 else if Tgap / 1> Pgap re_mq = i_mquant + 4 else re_mq = i_mquant + 1 where, Bm; Bit amount by mquant-1, B: Bit amount by mquant, Bp: Bit amount by mquant + 1, re_mq: real quantization stepsize 3. The method of claim 2, wherein the step (c3) uses a first code amount obtained by using a quantization step size larger than the reference quantization step size and a quantization step size one step smaller than the reference quantization step size. If the second code amount obtained by the reference quantization step size is smaller than the target code amount, not between the third code amounts obtained by the reference code amount, an actual quantization step size is obtained by subtracting an offset value from the reference quantization step size. How to. The method of claim 11, wherein the offset value is obtained by the following equation. Tgap = TB-B, Mgap = Bm-B if Tgap / 32> Mgap re_mq = i_mquant-8 else if Tgap / 16> Mgap re_mq = i_mquant-5 else if Tgap / 8> Mgap re_mq = i_mquant-3 else if Tgap / 4> Mgap re_mq = i_mquant-2 else re_mq = i_mquant-1 where, Bm; Bit amount by mquant-1, B: Bit amount by mquant, Bp: Bit amount by mquant + 1, re_mq: real quantization stepsize The method of claim 1, wherein the input image comprises slices composed of a predetermined number of macroblocks, and in step (a), the input image is discretely converted in units of a screen (intra picture) of the input image. The method of claim 1, wherein the actual quantization step size is one of macroblock units or slice units of several macroblocks. The method of claim 1, wherein the target code amount is a screen unit. The method of claim 1, (e) continuously and variable length encoding the compressed data to provide variable length encoded data; And and (f) generating the elementary stream by adding an additional header and outputting the variable length coded data as a bitstream. The method of claim 16, (g) extracting an elementary stream from the transmitted bitstream and removing an additional header from the elementary stream to provide a video stream; (h) variable length and continuous length decoding the video stream to provide continuous length decoded data; (i) inversely quantizing the continuous-length decoded data by the quantization step size used for encoding to provide inverse quantized data; And (j) inversely discrete-converting the inverse quantized data and outputting the original image data. In the apparatus for compression encoding and / or decoding the input video signal: A discrete conversion unit for performing discrete conversion of the input video signal to provide discrete conversion coefficients; An initial quantizer for quantizing the discrete transform coefficients using an initializing quantization matrix to provide initial quantized data; A code amount adjuster for providing a quantization step size for outputting a result closest to a target code amount among respective code amounts obtained by quantizing the initial quantized data in parallel using a plurality of quantization step sizes; And And a real quantizer for quantizing the initial quantized data by the actual quantization step size to provide compressed data. 19. The apparatus of claim 18, wherein the code amount regulator A plurality of quantizers for providing first to third quantized data by quantizing the initial quantized data in parallel using a reference quantization step size, one step larger quantization step size, and one step smaller quantization step size; A plurality of code length counters for counting the lengths of the first to third quantized data in predetermined size units to provide first to third code amounts; And An adjuster for providing a quantization step size for outputting a result closest to a target code amount among the first to third code amounts as an actual quantization step size and feedback input to the plurality of quantizers as an updated reference quantization step size. Device. 20. The apparatus of claim 19, wherein the adjuster comprises: a first code amount obtained by using a quantization step size that is one step larger than the reference quantization step size and a quantization step size that is obtained by using a quantization step size one step smaller than the reference quantization step size. When the second code amount obtained by the reference quantization step size is greater than the target code amount, the reference quantization step size is determined as the actual quantization step size, and when the second code amount approaches the target code amount. Device characterized in that changing the quantization step size of the edge of the screen less visually important. 21. The apparatus of claim 20, wherein the adjuster changes the quantization step size of the edge of the screen to a value one step larger than the determined reference quantization step size. 21. The apparatus of claim 20, further comprising a scan unit for scanning the initial quantized data from the edge of the screen, wherein the actual quantizer quantizes the initial quantized data to be scanned using a determined reference quantization step size, wherein the screen And the quantization step size of the edge uses a quantization step size that is larger than the determined reference quantization step size. 20. The apparatus of claim 19, wherein the adjuster comprises: a first code amount obtained by using a quantization step size that is one step larger than the reference quantization step size and a quantization step size that is obtained by using a quantization step size one step smaller than the reference quantization step size. If the second code amount obtained by the reference quantization step size is smaller than the target code amount, the reference quantization step size is determined as the actual quantization step size, and the second code amount is close to the target code amount. Device that varies the quantization step size from the center of the visually important screen until 24. The apparatus of claim 23, wherein the adjuster changes the quantization step size of the center portion of the screen to a value one step larger than the determined reference quantization step size. 24. The apparatus of claim 23, further comprising a scan unit for scanning the initial quantized data from a central portion of the screen, wherein the actual quantizer quantizes the initial quantized data to be scanned to a determined reference quantization step size, the center of the screen. And the quantization step size of the portion uses a quantization step size that is larger than the determined reference quantization step size. 20. The apparatus of claim 19, wherein the adjuster comprises: a first code amount obtained by using a quantization step size that is one step larger than the reference quantization step size and a quantization step size that is obtained by using a quantization step size one step smaller than the reference quantization step size. And an actual quantization step size in which an offset value is added to the reference quantization step size if the second code amount obtained by the reference quantization step size is not greater than three code quantities, and is larger than the target code amount. 27. The apparatus of claim 26, wherein the offset value is obtained by the formula below. Tgap = B-TB, Pgap = B-Bp if Tgap / 2> Pgap re_mq = i_mqaunt + 8 else if Tgap / 1> Pgap re_mq = i_mquant + 4 else re_mq = i_mquant + 1 where, Bm; Bit amount by mquant-1, B: Bit amount by mquant, Bp: Bit amount by mquant + 1, re_mq: real quantization stepsize 20. The apparatus of claim 19, wherein the adjuster comprises: a first code amount obtained by using a quantization step size that is one step larger than the reference quantization step size and a quantization step size that is obtained by using a quantization step size one step smaller than the reference quantization step size. An apparatus for providing an actual quantization step size obtained by subtracting an offset value from the reference quantization step size if the second code amount obtained by the reference quantization step size is smaller than the target code amount, not between three code quantities. . 29. The apparatus of claim 28, wherein the offset value is obtained by the following equation. Tgap = TB-B, Mgap = Bm-B if Tgap / 32> Mgap re_mq = i_mquant-8 else if Tgap / 16> Mgap re_mq = i_mquant-5 else if Tgap / 8> Mgap re_mq = i_mquant-3 else if Tgap / 4> Mgap re_mq = i_mquant-2 else re_mq = i_mquant-1 where, Bm; Bit amount by mquant-1, B: Bit amount by mquant, Bp: Bit amount by mquant + 1, re_mq: real quantization stepsize 19. The apparatus of claim 18, wherein the input image is composed of slices composed of a predetermined number of macroblocks, and the discrete conversion unit performs discrete conversion on a screen (intra picture) unit of the input image. 19. The apparatus of claim 18, wherein the actual quantization step size is one of macroblock units or slice units composed of several macroblocks. 19. The apparatus of claim 18, wherein the target code amount is a screen unit. The method of claim 18, A frame buffer which temporarily stores the initial quantized data and provides the actual quantizer while the code amount controller determines the actual quantization step size; A continuous-length and variable-length encoder that provides variable-length encoded data by performing continuous-length and variable-length encoding on the compressed data; And And a basestream encoder for generating the basestream by adding an additional header to the variable length coded data and outputting the basestream as a bitstream. The method of claim 33, wherein An elementary stream decoder for extracting an elementary stream from the transmitted bitstream and removing an additional header from the elementary stream to provide a video stream; A variable-length and continuous-length decoder configured to provide continuous-length decoded data by variable-length and continuous-length decoding the video stream; An inverse quantizer for inversely quantizing the continuous-length decoded data by a quantization step size used for encoding to provide inverse quantized data; And And an inverse discrete conversion unit for inverse discrete conversion of the inverse quantized data and outputting the original image data.