KR100601609B1 - Apparatus for decoding motion picture and method thereof - Google Patents

Abstract

Disclosed are a video decoding apparatus and a method for reducing memory capacity by using image compression. The present invention provides a demultiplexing process for separately separating a motion vector and a variable length code from an input bitstream, a decoding process for restoring the variable length code in macroblock units by a predetermined decoding algorithm, and a macro restored in the decoding process. Orthogonal transformation by dividing the data in block units into a plurality of blocks, and quantizing them, and then storing and storing them in a memory together with quantization level information. Process.

Description

Apparatus for decoding motion picture and method

1 is a block diagram showing a conventional moving picture decoding apparatus.

2 is a block diagram showing a moving picture decoding apparatus according to the present invention.

3 is a detailed diagram of the conversion encoder illustrated in FIG. 2.

4 is a detailed view of the transform decoder shown in FIG. 2.

5A to 5C illustrate an embodiment of a block partitioning method of a transform encoder.

6 shows a data structure multiplexed in the transform encoder.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture decoding apparatus and a method thereof, and more particularly, to a moving picture decoding apparatus and method for reducing memory capacity by decoding a moving picture using image compression.

Broadcasting environment based on the existing analog signal technology is rapidly developing into the proliferation of digital media, digital signal processing technology and digital communication technology. In particular, digital broadcasting such as high-definition television (hereinafter referred to as HDTV) broadcasting is rapidly spreading. But in order to receive HDTV, you need a dedicated digital TV receiver. One of the important factors that determine the price of such HDTV receivers is the memory required to receive HDTV-quality video. That is, 12-16 Mbytes or more of memory are required to decode the HDTV moving picture. In order to reduce such a large memory, moving picture decoding methods using an image compression method have been developed.

1 is a block diagram showing a conventional moving picture decoding apparatus.

First, the demultiplexer 110 distinguishes and outputs a variable length code for a discrete cosine transform (DCT) coefficient and a motion vector (MV) for motion compensation from an input bitstream. Variable Length Decoder (VLD) 120 restores the variable length code for the DCT coefficients to a quantization coefficient using a variable length decoding table (not shown). The inverse quantizer 130 restores the quantization coefficients to the DCT coefficients using the quantization table. Inverse Discrete Cosine Transform (IDCT) 140 generates data on prediction error through an inverse transform process on DCT coefficients. The motion compensator 150 receives the motion vector MV from the demultiplexer 110 and reads the predicted data block from the reference screen stored in the memory 190. In this case, the reference screen stored in the memory 190 is an image reduced according to a preset ratio compared to the original image. The shut-sampler 160 enlarges the block size by interpolating the prediction block data input from the motion compensator 150 to fit the block size in the original image. The multiplier 170 adds the macroblock coefficient output from the IDCT 140 and the output of the run-sampler 160. The downsampler 180 filters and then decimates the input data block to reduce the input macro block to a predetermined block size. The data blocks output from the downsampler 180 are block data reduced according to a preset ratio with respect to the input block, and these block data are written to the memory 190 and used for prediction or image display of the next frame.

As described above, the conventional technology decimates a restored block to reduce the memory capacity, reduces the block into a block having a predetermined size, and stores the same in a memory 190 and a reference screen stored in the memory 190 in a reduced form for motion compensation. Data is interpolated and restored to the original block size. This method reduces and stores the reconstructed image to a preset size, so that each pixel position in the image stored in the memory 190 is fixed, and block data stored at a specific position using a motion vector for motion compensation. Easy to search However, the image compression to reduce the memory capacity uses a very simple decimation method, so that the image quality of the restored video is very degraded. In particular, such image quality degradation is cumulative while the restoration of the moving image by the prediction method continues. have.

The technical problem to be achieved by the present invention is to restore a moving picture using a general moving picture reconstructor, and then to convert and store the compressed picture in a memory, and to convert and decode the data existing in the compressed form to compensate for motion. It is to provide a moving picture restoration apparatus to generate and its method.

In order to solve the above technical problem, the present invention provides a demultiplexing process for separating a motion vector and a variable length code from an input bitstream;

A decoding process of restoring the variable length code in macroblock units by a predetermined decoding algorithm;

A conversion encoding process of dividing the data of the macroblock unit reconstructed in the decoding process into a plurality of blocks, performing orthogonal transformation and quantization, and storing the data in a memory together with quantization level information;

And a transform decoding process of inversely quantizing and inversely orthogonally converting data stored in the memory to generate an output image.

In order to solve the above other technical problem, the present invention provides a video decoding apparatus,

Demultiplexing means for separating a motion vector and a variable length code from an input bitstream;

A memory for storing compressed data;

Decoding means for decoding the variable length code by a predetermined decoding scheme to restore data in macroblock units;

Conversion encoding means for compressing and encoding the macroblock data reconstructed by the decoding means in predetermined block units and stored in the memory;

Transform decoding means for decoding data stored in the memory by the transform encoding means to generate a prediction block;

And motion compensation means for compensating motion with the prediction block generated by the transform decoding means and the motion vector information.

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

2 is a block diagram showing a moving picture decoding apparatus according to the present invention, the demultiplexer 210, the VLD 220, the dequantizer 230, the IDCT 240, the motion compensator 250 operating in the same manner as the conventional one. In addition, a transform encoder 280 and a transform decoder 270 are added.

First, the demultiplexer 210 separates a motion vector and a variable length code from an input bitstream. The VLD 220, the inverse quantizer 230, and the IDCT 240, which are decoding means, perform variable length decoding, inverse quantization, and inverse orthogonal transformation, respectively, to restore data in macroblock units. The transform encoder 280 compresses and encodes the macroblock data to which the motion-compensated prediction block is added by the adder 260 in a predetermined block unit and stores them in the memory 290. The transform decoder 270 generates a prediction block according to the motion vector information from the macro block data stored in the memory 290 to generate an output image. The motion compensator 250 compensates for the motion in the prediction block generated by the transform decoder 270 with the input motion vector information.

3 is a detailed view of the conversion encoder 280 shown in FIG. 2.

First, the block generator 310 inputs macroblock data from the multiplier 260 and divides the block into a plurality of small blocks. Some embodiments of block division are illustrated in FIGS. 5A-5C, and can be modified in various other forms. When the block generation unit 310 changes the shape of a block that is adaptively divided for each macro block, pattern information used for the block division is generated. On the other hand, when the block is divided into fixed forms, the pattern information is not generated.

The transformer 320 performs orthogonal transformation on the block generated by the block generator 310. In this case, as an example of the orthogonal transformation, DCT which is typically used for image compression may be used.

The first controller 340 generates a control signal for bit allocation for each block generated by the block generator 310 using the macro block data input from the adder 260 and transmits the generated control signal to the quantizer 330. . That is, the first controller 340 determines the quantization level of the transform coefficients of each block generated by the transform unit 320 using the characteristics of each block. The quantizer 330 performs quantization on each of the transform coefficients using the quantization level input from the first controller 340. The multiplexer 350 includes quantization coefficients generated by the quantizer 330 and quantization level data for each block generated by the first controller 340 or data about a block pattern generated by the block generator 310. 6 is multiplexed as shown in FIG. 6 to be written to the memory 290. As shown in FIG. 6, data is divided into a header area and a data area (first block data ... Nth block data), and the header area includes a quantization level indicating bit allocation of each block. Information about the block pattern or information about the block pattern is recorded, and the quantization coefficient of each block is recorded in the data area (first block data ... N-th block data).

4 is a detailed view of the transform decoder 270 shown in FIG.

The demultiplexer 450 calculates a data area selected for motion compensation using the input motion vector, and reads compressed macro block data belonging to the selected area from the memory 290. The compressed macro block data stored in the memory 290 is divided into a header area including a quantization level or a block division pattern and a data area in which a transform coefficient is recorded. The demultiplexer 450 applies a quantization level to the second controller 440, applies block information to the prediction block generator 410, and applies transform coefficients for the block to the dequantizer 430. The second controller 440 analyzes the header information input from the demultiplexer 450, generates a level signal necessary for dequantization of each block, and applies it to the dequantizer 430. The inverse quantizer 430 dequantizes the quantization coefficients for each block input from the demultiplexer 450 using the quantization level and then generates transform coefficients for each block. The inverse transformer 420 generates block data by performing an inverse orthogonal transformation process corresponding to the orthogonal transformation. The side block generator 410 receives the motion vector, the block information, and the block data, and selects data necessary for motion compensation.

In addition, in the present invention, the following constraints may be considered for the optimal compression / restore effect. That is, the input and output of the memory is made in units of words, and one word is usually composed of 1-8 bytes. Using such an input / output unit, the conversion encoder of FIG. 3 adjusts the total bit generation amount for a macro block in word units, and adjusts the quantization level of each block to change the bit generation amount for each block in word units. Therefore, when reading the compressed macro block data in the transform decoder of FIG. 4, redundant data due to mismatch of word boundaries is greatly reduced.

As described above, according to the present invention, a very high compression ratio is obtained when storing a reconstructed image in a memory through transform coding that is effective for image encoding and adaptive bit allocation for each transform block, compared to the prior art based on decimation and interpolation. It can achieve and also can improve the picture quality of the moving picture which is restored compared with the conventional.

Claims (7)

The reconstructed macroblock in the decoder which separates the motion vector and the variable length code from the input bitstream into macroblock units by variable length decoding, inverse quantization, and inverse orthogonal transformation to perform motion compensation. A moving picture decoding method for efficiently storing unit data and generating a prediction block for motion compensation, Orthogonal transformation by dividing the restored macroblock unit data into a plurality of blocks, quantizing the orthogonal transformed data to a predetermined quantization level, multiplexing the quantized data and the quantization level information, and storing the data in a predetermined memory region. Transcoding process; Inverse quantization of data stored in the predetermined memory region with reference to a predetermined quantization level, inverse orthogonal transformation of the inverse quantized data to generate block data, and applying the motion vector to the generated block data to perform the motion. A video decoding method comprising a transform decoding process for generating compensation block data for compensation. The method of claim 1, wherein the transform encoding process comprises: dividing the macro block data into a plurality of blocks; Generating transform coefficients by orthogonally transforming the divided blocks; Determining a quantization level with reference to the macro block data; Quantizing the transform coefficients generated in the process at the quantization level determined in the process; And multiplexing the quantized transform coefficients, the quantization level, and the block partitioning information into blocks in a unit of memory in the memory. 3. The moving picture decoding method according to claim 2, wherein the quantization level is determined such that the information amount is generated in a predetermined number of bytes. The method of claim 1, wherein the transform decoding process comprises: demultiplexing block data of a specific region stored in the process using the motion vector to provide a quantization coefficient, a quantization level, and block information; Inversely quantizing the quantization coefficients with reference to the demultiplexed dequantization level in the process and then performing inverse orthogonal transformation on the coefficients; And generating a prediction block using the inverse transformed coefficient, the block information, and the motion vector in the above process. The reconstructed macroblock in the decoder which separates the motion vector and the variable length code from the input bitstream into macroblock units by variable length decoding, inverse quantization, and inverse orthogonal transformation to perform motion compensation. A moving picture decoding apparatus for efficiently storing unit data and generating a prediction block for motion compensation, Memory means for storing compressed data; The restored macroblock unit data is divided into orthogonal transforms into a plurality of blocks, the orthogonal transformed data is quantized to a predetermined quantization level, and the quantized data and the quantization level information are multiplexed to a predetermined region of the memory means. Conversion encoding means for storing; Inverse quantization of data stored in a predetermined area of the memory means with reference to the quantization level, inverse orthogonal transformation of the inverse quantized data to generate block data, and the motion by applying the motion vector to the generated block data And a video decoding means for generating the compensation predictive block data. The conversion encoding means according to claim 5, A transformer for orthogonally converting the restored macroblock into a plurality of blocks; A first controller configured to determine bit allocation for each transform block according to characteristics of the restored macroblock; A quantizer for quantizing the coefficients of the transform block according to the bit allocation determined by the first controller; And a multiplexer which multiplexes the quantized coefficients and the bit allocation information by the quantization unit. 6. The conversion decoding unit according to claim 5, wherein A demultiplexer which demultiplexes a quantization coefficient, bit allocation information, and block information by reading data of a specific position stored in the memory using the motion vector; A second controller configured to determine a dequantization level by referring to bit allocation information demultiplexed by the demultiplexer; An inverse quantizer for inversely quantizing a quantization coefficient according to the inverse quantization level determined by the second controller; An inverse transformer for inverse orthogonal transforming the inverse quantized coefficients in the inverse quantization unit; And an edge block generator for generating a prediction block based on the inverse transform coefficient, the block information, and the motion vector.

Images