KR20050115779A - Apparatus and method for encoding video signal based on multi-resolution - Google Patents

Abstract

The present invention relates to a video encoding apparatus and a method for introducing a multi-resolution motion estimation algorithm. The present invention stores the low resolution macroblock information generated in the pre-processing step and the pre-processing step to generate the hierarchical structure in macro units by filtering and subsampling the input image signals, and generates the pre-processing step at the same time. And a video encoding step of performing motion compensation based on the original resolution macro information corresponding to the current frame, the low resolution macro information, and the original resolution image and the low resolution level image information of the previous reference frame stored in the memory.

Description

Apparatus and method for encoding video signal based on multi-resolution

The present invention relates to a moving picture coding system, and more particularly, to a moving picture coding apparatus and method using a multi-resolution motion estimation algorithm.

In general, the portion of the video encoder, such as MPEG, takes up the most amount of motion is motion estimation. Especially, in SDTV or HDTV video encoder, the search area should be wide, which is very complicated in software and hardware implementation. Many fast motion search techniques have been developed to solve this problem. Among them, motion estimation techniques based on multi-resolution frame structure are advantageous in terms of performance or ease of hardware implementation. Therefore, multi-resolution-based motion estimation algorithms and hardware architectures have been developed. Literature data applying such a multi-resolution motion estimation technique is described in detail in "Korean Patent Application No. P2002-0041985". This multi-resolution motion estimation technique has the advantages of high performance and low computational capacity, but also has the disadvantage of storing low resolution images in memory and increasing memory I / O bandwidth when accessing them. .

1 is a block diagram of a general video encoder system employing a conventional multi-resolution motion estimation algorithm.

FIG. 1 illustrates a preprocessor 120 generating a low resolution image, an external frame memory 140 storing original images and low resolution images output from the preprocessor 120, and current frame information from an external frame memory 140. The video encoder 130 reads and compresses the reference frame information, and a bus between each block and the external frame memory 140.

First, a multi-resolution motion estimation algorithm is described in "Korean Patent Application No. P2002-0041985". Referring to this document, let the number of resolution levels be 3 (L = 3). First, the input image data is written to the external frame memory 140 through the buffer 110 (WR_L0_frame). The preprocessor 120 reads the image data RD_L0_frame from the external frame memory 140 again. The preprocessor 120 performs low pass filtering (LPF) and generates a low resolution image through sub-sampling. The low resolution images L1_frame and the low resolution images L2_frame, which are low resolution images output from the preprocessor 120, are stored in the external frame memory 140 through a bus. This is referred to as WR_L1_frame and WR_L2_frame, respectively. The video encoder 130 performs motion estimation by reading current frame information and previous reference frame information. In this case, motion estimation is performed in units of macroblocks (16x16). That is, the 16x16 macroblock and the corresponding search region information in the current frame are read in the previous reference frame. The original resolution current frame image information is cur_L0_frame, the middle resolution level current frame image information is cur_L1_frame, and the lowest resolution level current frame image information is expressed as cur_L2_frame. Meanwhile, in the case of the previous reference frame information, the original reference frame image information is referred to as ref_L0_frame, the middle level image is referred to as ref_L1_frame, and the lowest resolution level image is referred to as ref_L2_frame. The video encoder 130 reads the information from the external frame memory 140 in order to perform L2 search, L1 search, and L0 search. Video encoder 130 is buffered (130-8) and motion estimation (130-9) and compensation (130-10), DCT (130-1), quantization (130-2), VLC (130-3), buffering The bitstream is generated through 130-4. In addition, in the case of a P-frame, the reconstructed image WR_rec_L0_frame is external quantized through the inverse quantization 130-5, IDCT 130-6, and buffering 130-7 to compensate for motion of other frames. Should be stored in As shown in FIG. 1, the conventional video encoder needs to read two or more L0_frames, L1_frames, and L2_frames from the external frame memory 140 on a P-frame basis, and L0_ corresponding to one frame. Frames, L1_frames, and L2_frames should be written to the external frame memory 140, and a reconstructed L0_frame corresponding to one frame should be written to the external frame memory 140. Therefore, the video encoding apparatus using the conventional three-level multi-resolution motion estimation technique has a problem of requiring a large memory input / output bandwidth. In addition, there is a problem in that low-resolution image information corresponding to two frames is additionally stored in a memory.

SUMMARY OF THE INVENTION The present invention has been made in an effort to provide a multi-resolution video coding method and system capable of reducing memory storage capacity and memory input / output bandwidth by generating a low resolution video inside a video encoder to which a multi-resolution motion estimation technique is applied. have.

In order to solve the above technical problem, the present invention provides a moving picture coding unit for encoding an input video signal, the buffer unit and a video unit of the moving picture encoding unit in the video system having a memory unit for storing the video signal of the video encoding unit In the compression method,

A pre-processing step of generating the input image signals into a hierarchical structure in macro units through filtering and subsampling;

The low resolution macroblock information generated by the preprocessor is stored in the memory unit, and the original resolution macro information corresponding to the current frame generated in the preprocessing step and the low resolution macro information and the previous reference frame stored in the memory unit. And a video encoding step of performing motion compensation based on the original resolution image and the low resolution level image information.

In order to solve the above other technical problem, the present invention provides a video encoding system,

A buffer unit for buffering an input video signal;

A pre-processing unit generating a hierarchical structure in macroblock units by filtering and subsampling the video signal input from the buffer unit;

A memory unit for storing the low resolution level image signal and the original resolution signal generated by the preprocessor in block units, and storing the original resolution image and the low resolution level image information of a previous reference frame;

A video for performing motion compensation based on the original resolution image information and the low resolution level image information corresponding to the current frame generated by the preprocessor and the original resolution image and the low resolution level image information of the previous reference frame stored in the memory. Characterized in that it comprises a sign.

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

2 is a first embodiment of a video encoding system according to the present invention.

The video system of FIG. 2 corresponds to a buffer 210 for buffering an input image, a preprocessor 220 for receiving hierarchical frame information from the buffer 210, and a preprocessor 220 for generating a hierarchical image. The video encoder 230 and the preprocessor 220 that receive the low resolution image information as well as the original resolution image and the low resolution image information of the previous reference frame and compress the image with the predetermined compression algorithm. External frame memory 240 for storing the low resolution images (L1_frame and L2_frame) of the current frame, and also for storing the original and low resolution images of the previous reference frame, and for reconstructing the reconstructed image in the case of a P-frame, and between them. Equipped with a bus. The video encoder 230 may include a motion estimator 330-9 and a motion compensator 3330-10, a DCT 330-1, a quantizer 330-2, a VLC 330-3, and an inverse quantizer ( 330-5) and IDCT device 330-6.

Referring to FIG. 2, the input image is input to the preprocessor 220 via the buffer 210. The preprocessor 220 generates low-resolution images of the current frame in units of macroblocks (MB) through LPF and sub-sampling. For example, the image of the highest resolution level is the original image, and the image of the intermediate resolution level is 1/2 the size of each of the width and length of the original image. The image of the lowest resolution level is 1/4 the horizontal and vertical size of the image of the intermediate resolution level. The intermediate level current image information L2_MB and the lowest resolution current image information L1_MB generated by the preprocessor 220 are stored in the external frame memory 240 in units of macroblocks. At the same time, the original resolution image information L0_MB, L2_MB, and L1_MB of the current frame generated by the preprocessor 220 are transmitted to the video encoder 230. In the case of a P-frame or a B-frame, the video encoder 230 reads ref_L0_SW, ref_L1_SW, ref_L2_SW, which are previous reference frames stored in the external frame memory 240, and performs encoding through motion compensation. SW is a search window. That is, search areas for level-specific searching are read from the external frame memory 240. The video encoder 230 reads the information from the external frame memory 240 in order to perform L2 search, L1 search, and L0 search to perform motion vector estimation 230-9 and motion compensation 230-10. Thereafter, the bitstream is generated through the DCT 230-1, the quantization 230-2, the VLC 230-3, and the buffer 230-4. Also, in the case of a P-frame, the macroblock reconstructed through the inverse quantization 230-5 and the IDCT 230-6 in the video encoder 230 passes through the buffer 230-7 to compensate for the motion of the next frame. It is stored in the memory 240. The operation of each frame is as follows.

1) When input video is I-frame

The input image of the I-frame is stored in the external frame memory 240 in units of macroblocks in order to estimate the motion in the next frame through L2_MB and L1_MB of the frame via the preprocessor 220. Since the I-frame has no motion estimation and compensation, only its original resolution block (L0_MB) is transferred to the video encoder 230, and the reconstructed image block (L0_MB) generated during encoding is the external frame memory 240 for motion compensation of the following frames. )

2) When input video is P-frame

The input image of the P-frame is stored in the external frame memory 240 through the preprocessor 220 to L2_MB and L1_MB of the frame for motion estimation in subsequent frames. At the same time, the original resolution image (cur_L0_MB), cur_L1_MB, and cur_L2_MB of the current frame are transmitted to the video encoder 230 as it is for motion estimation. Meanwhile, the information of the search region information ref_L0_SW, ref_L1_SW, and ref_L2_SW in the previous reference frame is read from the external frame memory 240 for motion estimation for each of L2, L1, and L0 levels. Meanwhile, the reconstructed block information of the current block generated during the encoding process is stored in the external frame memory 240 to compensate for the motion of the next frame.

3). When input video is B-frame

The input image of the B-frame is generated through the preprocessor 220 and L2_MB and L1_MB of the frame, and cur_L1_MB and cur_L2_MB are left in the video encoder 230 together with the original resolution image (cur_L0_MB) of the current frame for motion estimation. Delivered. At this time, since the B-frame is not used as a reference frame, cur_L1_MB and cur_L2_MB are not stored in the external frame memory 240. Meanwhile, the information of the search region information ref_L0_SW, ref_L1_SW, and ref_L2_SW in the previous reference frame is read from the external frame memory 240 for motion estimation for each of L2, L1, and L0 levels. In this case, since the bidirectional case should be checked, the video encoder 230 inputs the ref_L0_SW, ref_L1_SW, and ref_L2_SW information in two directions. In this case, since the B-frame is not used as the reference frame, the reconstructed block information of the current block is not stored in the external frame memory 240 during the encoding process.

3 is a second embodiment of a video encoding system according to the present invention.

Referring to FIG. 3, the preprocessor 320 and the video encoder 330 include a video encoding apparatus 300. The input image is directly stored in the external frame memory 340 via the buffer 310. The preprocessor 320 receives the current macroblock and generates low resolution images of the current frame in units of macroblocks (MBs) through LPF and sub-sampling. That is, the intermediate level current image information L2_MB and the lowest resolution current image information L1_MB generated by the preprocessor 320 are stored in the external frame memory 340 in units of macroblocks. At the same time, the original resolution image information L0_MB, L2_MB, and L1_MB of the current frame are transmitted to the video encoder 330. In the case of a P-frame or a B-frame, the video encoder 230 reads ref_L0_SW, ref_L1_SW, ref_L2_SW, which are previous reference frames, and performs encoding through motion compensation. That is, search areas for level-specific searching are read from the external frame memory 340. The video encoder 330 reads the information from the external frame memory 340 in order and performs L2 search, L1 search, and L0 search to perform motion vector estimation 330-9 and motion compensation 330-10. Thereafter, the bitstream is generated through the DCT 330-1, the quantization 330-2, the VLC 330-3, and the buffer 330-4. Also, in the case of an I-frame / P-frame, the macroblock reconstructed through the inverse quantization 330-5 and the IDCT 330-6 in the video encoder 330 is buffered to compensate for the motion of the next frame. It is stored in the external frame memory 340 via. The operation of each frame is as follows.

1. When the input image is I-frame

The input image is first stored in the external frame memory 340 via the buffer 310. The preprocessor 320 reads the current block (cur_L0_MB) from the external frame memory 340 to generate corresponding L2_MB and L1_MB, and to the external frame memory 340 in macroblock units for motion estimation in subsequent frames. Output Since the I-frame has no motion estimation and compensation, only its original resolution block (L0_MB) is transferred to the video encoder 330, and the reconstructed image block (L0_MB) generated during encoding is performed using an external frame memory ( 340.

2. When input video is P-frame

The input image is first stored in the external frame memory 340 via the buffer 310. The preprocessor 320 reads the current block (cur_L0_MB) from the external frame memory 340 to generate corresponding L2_MB and L1_MB, and to the external frame memory 340 in macroblock units for motion estimation in subsequent frames. Output At the same time, the original resolution image (cur_L0_MB), cur_L1_MB, and cur_L2_MB of the current frame are transmitted to the video encoder 330 as it is for motion estimation. On the other hand, the information of the search region ref_L0_SW, ref_L1_SW, and ref_L2_SW in the previous reference frame is read from the external frame memory 340 for motion estimation for each of L2, L1, and L0 levels. Meanwhile, the reconstructed block information of the current block generated during the encoding process is stored in the external frame memory 340 for motion compensation of the next frame.

3. When input video is B-frame

The input image is first stored in the external frame memory 340 via the buffer 310. The preprocessor 320 reads the current block (cur_L0_MB) from the external frame memory 340 to generate corresponding L2_MB and L1_MB, and cur_L1_MB and cur_L2_MB are displayed together with the original resolution image (cur_L0_MB) of the current frame for motion estimation. Passed to the encoder 330 as it is. At this time, since the B-frame is not used as a reference frame, cur_L1_MB and cur_L2_MB are not stored in the external frame memory 340. On the other hand, the information of the search region ref_L0_SW, ref_L1_SW, and ref_L2_SW in the previous reference frame is read from the external frame memory 340 for motion estimation for each of L2, L1, and L0 levels. At this time, since the bidirectional case should be checked, information of the two directions of ref_L0_SW, ref_L1_SW, and ref_L2_SW is input to the video encoder 330. In this case, since the B-frame is not used as the reference frame, the reconstructed block information of the current block is not stored in the external frame memory 340 during the encoding process.

The present invention is not limited to the above-described embodiment, and of course, modifications may be made by those skilled in the art within the spirit of the present invention.

The present invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, hard disk, floppy disk, flash memory, optical data storage device, and also carrier waves (for example, transmission over the Internet). It also includes the implementation in the form of. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

As described above, according to the present invention, by providing a preprocessor for generating a low resolution image inside the video encoder, not only the size of the memory but also the bandwidth can be greatly reduced. In addition, since LPF and sub-sampling are performed in units of macro blocks in the preprocessor, the hardware size is significantly reduced compared to the prior art using line memory.

1 is a block diagram of a general video encoder system employing a conventional multi-resolution motion estimation algorithm.

2 is a first embodiment of a video encoding system according to the present invention.

3 is a second embodiment of a video encoding system according to the present invention.

Claims (6)

In a moving picture compression method of a moving picture coder in a moving picture system having a buffer unit for buffering an input image, an external memory unit for storing a video signal, and a moving picture coder, A pre-processing step of generating the input image signals into a hierarchical structure in macro units through filtering and subsampling; The low resolution macroblock information generated in the preprocessing step is stored in the external memory unit, and the raw resolution macro information and the low resolution macro information corresponding to the current frame generated in the preprocessing step are inputted and stored in the external memory unit. And a video encoding step of performing motion compensation together with the original resolution image and the low resolution level image information of the previous reference frame. The method of claim 1, wherein the moving picture encoding step receives raw resolution block information generated in the preprocessing step when the input video signal is an I-frame, and raw resolution macro information of the current frame generated in the preprocessing step when the input video signal is an I-frame. And performing motion compensation based on low resolution macro information and search region information for each level in a previous reference frame stored in the external memory unit, and in the case of a B-frame, original resolution macro information and low resolution macro information of the current frame generated in a preprocessing step. And motion compensation based on the search area information for each level in the bidirectional previous reference frame stored in the external memory unit. In a moving picture compression method of a moving picture coder in a moving picture system having a buffer unit for buffering an input image, an external memory unit for storing a video signal, and a moving picture coder, Storing the input image signals in the external memory through the buffer unit; A pre-processing step of generating an image signal of a macroblock unit among the image signals stored in the external memory in a hierarchical structure through filtering and subsampling; The low resolution macroblock information generated in the preprocessing step is stored in the external memory unit, and the raw resolution macro information and the low resolution macro information corresponding to the current frame generated in the preprocessing step are inputted and stored in the external memory unit. And a video encoding step of performing motion compensation together with the original resolution image and the low resolution level image information of the previous reference frame. The video encoding method according to claim 3, wherein the preprocessing step comprises reading a current macroblock from the external memory to generate a corresponding low level macroblock. In the video encoding system, Buffer means for buffering an input video signal; The image signal input from the buffer means is generated in a hierarchical structure in units of macroblocks through filtering and subsampling, and original resolution macroblock information corresponding to the generated current frame and macroblock information of a low resolution level and previously stored Video encoding means for performing motion compensation based on the original resolution image information of the reference frame and the low resolution level image information; And a memory means for storing the low resolution level image signal and the original resolution signal generated by the video encoding means in block units, and storing the original resolution image and the low resolution level image information of the previous reference frame. . The video encoding means according to claim 5, A pre-processing unit generating a hierarchical structure in macroblock units by filtering and subsampling the video signal input from the buffer unit; Based on the original resolution macro block information corresponding to the current frame generated by the preprocessor and the low resolution level macro block information, the original resolution image information of the previous reference frame stored in the memory and the low resolution level image information, And a moving picture encoding unit for estimating.