KR101074919B1 - Apparatuses and methods of decoding distributed video coded video - Google Patents

Abstract

An apparatus and method for decoding distributed video encoded video using variable block motion prediction are provided. The apparatus for decoding a distributed video encoded image includes a key picture decoder for reconstructing a key picture transmitted from the encoder and a size of a block for performing motion prediction using the reconstructed key picture to generate and generate a motion vector. A motion vector generator for outputting a motion vector, an auxiliary information generator for generating auxiliary information using the reconstructed key picture and the generated motion vector, a parity bit transmitted from the encoding apparatus, and the auxiliary information And a channel code decoding unit for estimating a quantized value, and an image restoring unit for restoring a current picture to be decoded using the quantized value and the auxiliary information estimated by the channel code decoding unit. Accordingly, by performing the motion prediction process performed at the time of generating the auxiliary information variably to generate the auxiliary information well-removed, the performance of the decoder can be improved by reducing the amount of parity bits transmitted from the encoding apparatus.

Description

Apparatus and method for decoding distributed video encoded image {APPARATUSES AND METHODS OF DECODING DISTRIBUTED VIDEO CODED VIDEO}

The present invention relates to an apparatus for decoding an image and a method thereof, and more particularly, to an apparatus and a method for decoding a distributed video encoded image.

MPEG, H.26x, and other compression standards are widely used as efficient compression technologies for video players, customized video information services (VOD), video telephony, digital multimedia broadcasting (DMB), and video transmission in wireless mobile environments. The compression standards have a large gain in coding efficiency by eliminating temporal redundancy. As a representative method for reducing the temporal redundancy, there are motion prediction and compensation techniques. However, since the motion prediction and compensation technique requires a relatively large amount of computation in the video encoder, power consumption increases. Therefore, in a limited resource environment such as a sensor network, in order to reduce the power of the encoder, reducing the complexity of the encoder has emerged as an important technical problem.

As one of the methods for solving the complexity problem of the encoder, a Distributed Video Coding (DVC) method based on the Weiner-Ziv theory has been in the spotlight. This distributed video encoding technique can reduce the amount of computation of the encoder because motion prediction and compensation can be performed in the decoder without loss of code and gain.

According to the distributed video coding technique based on the Weiner-Jib theory, the decoder generates auxiliary information about the current picture using similarity between neighboring pictures, and the auxiliary information is that the virtual channel noise is added to the current picture to be reconstructed. In this case, the current picture is reproduced by receiving a parity bit generated as a channel code from an encoder and removing noise in the auxiliary information.

1 is a diagram illustrating a configuration of an encoder 110 and a decoder 130 corresponding to the conventional Wyner-Ziv coding technique.

The encoder 110 according to the Wyner-Ziv coding technique classifies pictures of source video content into two types. One is a picture to be encoded by using a distributed video coding method (hereinafter referred to as a 'WZ picture'), and the other is a picture to be encoded by a conventional coding method instead of a distributed video coding method (hereinafter, referred to as a 'key picture'). )to be.

The key pictures are encoded by the intra picture method of, for example, H.264 / AVC in the key picture encoder 114 and transmitted to the decoder 130. The key picture decoder 133 of the decoder 130 corresponding to the encoder 110 according to the conventional Wyner-Ziv coding technique reconstructs the transmitted key pictures. The auxiliary information generating unit 134 generates side information corresponding to the WZ picture by using the key picture restored by the key picture decoding unit 133 and sends the auxiliary information to the channel code decoding unit 131. Output

The auxiliary information generator 134 assumes linear motion between key pictures located before and after the current WZ picture, and generates side information corresponding to the WZ picture to be reconstructed using interpolation. In some cases, the interpolation method may be used instead of the interpolation method, but in most cases, the interpolation method is used because the noise in the auxiliary information generated by the interpolation method is smaller than the noise in the auxiliary information generated by the interpolation method. In order to encode, the quantization unit 112 of the encoder 110 performs quantization on the WZ picture and outputs the quantization value of the WZ picture to the block unitization unit 112. The block unit unit 111 classifies the quantized value of the input WZ picture into predetermined coding units. The channel code encoder 113 generates a parity bit for each coding unit by using the channel code.

The generated parity bits are temporarily stored in a parity buffer (not shown), and are sequentially transmitted when the decoder 130 requests parity through a feedback channel. The channel code decoder 131 of FIG. 1 receives the parity transmitted from the encoder 110 and estimates the quantized value. The image restorer 132 of FIG. 1 receives the quantized value estimated by the channel code decoder 131 and inversely quantizes the quantized value to restore the WZ picture.

The distributed video encoding method based on the Wyner-Ziv theory is a technique for reconstructing the current WZ picture by correcting the noise added to the auxiliary information generated by the decoder using parity. Therefore, the less noise added to the generated auxiliary information, the smaller the amount of parity required. Therefore, it is important to generate auxiliary information well without noise in order to have good performance in terms of rate distortion.

The conventional auxiliary information generating method estimates a motion vector using a fixed size block (eg, 8x8) existing in the reconstructed key picture and reconstructs the motion vector of the auxiliary information to be reconstructed in consideration of the distance between pictures. Obtained from motion vectors between key pictures. The decoding unit in the key picture indicated by the motion vector of the auxiliary information thus obtained is generated as auxiliary information.

2 illustrates a conventional generation of auxiliary information using an interpolation method. The process of obtaining the motion vector of the auxiliary information to be reconstructed from the motion vector obtained by using a fixed size block between the reconstructed key pictures. Considering the distance between the pictures, the motion vector of the auxiliary information is determined by the motion vector of the key pictures. It can be seen that 1/2.

3 illustrates a conventional generation of auxiliary information using extrapolation. The process of obtaining the motion vector of the auxiliary information to be reconstructed from the motion vector obtained by using a fixed size block between the reconstructed key pictures. Considering the distance between the pictures, the motion vector of the auxiliary information is determined by the motion vector between the key pictures. It can be seen that the same.

As such, when the motion vector of the auxiliary information to be generated is obtained from the motion vector using a fixed size block existing in the reconstructed key picture, the motion is complicated between pictures or the motion does not change linearly, or an object And when the background suddenly disappears or appears, it may generate incorrect auxiliary information. In particular, when the size of the fixed block used for motion prediction between key pictures increases, the accuracy of the obtained motion vector is similar to the motion of an actual image in a simple region such as a background, but not in a part where a complex object exists.

In contrast, when the size of the fixed block used for motion prediction between key pictures is reduced, the number of pixels used for estimating the motion vector is reduced, so that the motion vector obtained through the motion prediction between the key pictures is a real image. Frequent occurrences of inconsistencies with the movements of For this reason, 8x8 is generally used for the block size used for motion prediction for generating auxiliary information.

As in the above case, since the auxiliary information generated by the motion prediction using the fixed block size has a lot of noise, the transmitted parity cannot sufficiently remove the noise.

Accordingly, an object of the present invention is to provide a distributed video decoding apparatus using variable block motion prediction.

Another object of the present invention is to provide a distributed video decoding method using variable block motion prediction.

In accordance with one aspect of the present invention, there is provided a apparatus for decoding a distributed video encoded image, the apparatus comprising: a key picture decoder configured to reconstruct a key picture transmitted from an encoding apparatus; A motion vector generator configured to determine a size of a block for performing motion prediction using the reconstructed key picture, generate a motion vector, and output the generated motion vector; An auxiliary information generator configured to generate auxiliary information using the reconstructed key picture and the generated motion vector; A channel code decoder for estimating a quantized value using the parity bits transmitted from the encoding apparatus and the auxiliary information; And an image restoring unit for restoring a current picture to be decoded by using the quantized value estimated by the channel code decoding unit and the auxiliary information.

According to the present invention, the size of a block for motion prediction between key pictures is variably applied based on a predetermined criterion when generating auxiliary information used in the decoding process of distributed video encoding.

Therefore, by more precisely estimating the motion vector between key pictures according to the present invention, the performance of the reconstructed image can be significantly improved.

That is, the present invention estimates the motion vector while variably applying the size of the block existing in the reconstructed key picture to the motion vector of the auxiliary information to be generated, thereby making the motion vector more sophisticated than the motion vector estimated using the conventional fixed size. The vector can be obtained to reduce the noise of the generated auxiliary information to improve the quality of the reconstructed image.

1 is a block diagram illustrating a configuration of an encoder and a decoder corresponding thereto according to a conventional Weiner jib coding technique.
2 is an exemplary diagram for explaining conventional generation of auxiliary information using an interpolation method.
3 is an exemplary diagram for explaining conventional generation of auxiliary information using extrapolation.
4 is a block diagram illustrating a configuration of an apparatus for decoding a distributed video encoded image according to an embodiment of the present invention.
5 is a view for explaining a process of generating a motion vector according to an embodiment of the present invention.
6 is a diagram illustrating a configuration of a motion vector generating unit according to an embodiment of the present invention.
7 is a flowchart illustrating a motion vector generation process according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Specific details appear in the following description, which is provided to help a more general understanding of the present invention, and it is common knowledge in the art that such specific matters may be changed or modified within the scope of the present invention. It is self-evident to those who have.

4 shows a wine-jib encoding and decoding system according to an embodiment of the present invention.

The Wyner-Ziv encoding and decoding system 300 according to an embodiment of the present invention is a distributed video encoded image decoding apparatus 300 including a function of generating auxiliary information through motion estimation using a wiener-jib encoding apparatus 10 and a variable block. I include)

Referring to FIG. 4, the wine-jib encoding apparatus 10 includes a WZ picture encoder 11 and a key picture encoder 12.

The wine-jib encoding apparatus 10 divides pictures of the source video content into wine-jib pictures and key pictures according to the wine-jib coding scheme. For example, even-numbered pictures of the source video content may be selected as key pictures or odd-numbered pictures may be selected as a wine-jib picture.

The wine-jib encoding apparatus 10 encodes the wine-jib pictures and the key pictures and provides them to the decoding apparatus 300.

In addition, the decoding apparatus 300 according to the present invention includes a key picture decoder 320, a channel code decoder 310, an image restorer 330, an auxiliary information generator 340, and a block size determiner 350. And a picture buffer 360.

The key picture decoder 320 reconstructs the key picture using the information received from the key picture encoder 12.

The motion vector generator 350 determines a size of a block to be used for motion prediction by using the reconstructed key picture.

The auxiliary information generator 340 generates auxiliary information on the current WZ picture to be reproduced using the reconstructed key picture and the motion vector generated by the motion vector generator 350.

The channel code decoder 310 estimates the quantized values using the auxiliary information received from the auxiliary information generator 340 and the parity bits received from the wine-jib encoding apparatus 10.

If it is determined that the channel code decoding unit 310 cannot estimate a reliable quantized value when performing channel code decoding, the channel code decoding unit 310 continuously performs the Wier-Jib with the parity bit within a predetermined limit until a reliable estimation is possible. It may be configured to request and receive the encoding device 10.

In this case, the wine-jib encoding apparatus 10 is efficient in terms of rate-distortion performance by receiving only the amount of parity necessary for decoding from the wine-jib encoding apparatus 10. This is possible when there is a reverse channel (ie a feedback channel) requesting parity bits.

In order to alleviate this problem, the channel code decoding unit 310 is configured to receive a predetermined amount of parity bits in advance without requesting a parity request at a time, so as not to request parity on the reverse channel while using the same. Can be.

Also, in the above case, after exhausting the sent parity bits, if it is determined that the reliability is still low, the decoding device 300 may be configured to further request the parity bits to the encoding device 10. In addition, assuming that the reverse channel is not used, the Weiner-Jib encoding apparatus 10 transmits a predetermined amount of parity calculated or set in advance to the decoding apparatus 300, and the decoding apparatus 300 is configured to not require a parity bit. May be

In addition, the channel code used in the channel code decoding unit 310 of FIG. 4 may use a turbo code that is found to almost reach the Shannon limit or may use an LDPC channel code. It is also apparent that other channel codes with good coding efficiency and error correction can be used.

The image restorer 330 restores the current WZ picture using the quantized value and the auxiliary information estimated by the channel code decoder 31. The auxiliary information is generated by the auxiliary information generator 340.

The auxiliary information generator 340 generally uses an interpolation method that assumes linear change between pictures in generating auxiliary information. In this case, the change between the video pictures is caused by the movement of the object, the movement of the camera, the exposure area and the light. Except for changes in the exposure area and light, the difference between these pictures corresponds to the movement of pixels between the pictures. If the motion vector of the pixel can be known accurately, it is possible to accurately predict the pixels of the picture to be interpolated.

That is, the auxiliary information generator 340 generates auxiliary information based on the motion vector provided from the motion vector generator 350.

The auxiliary information generator 340 uses side information corresponding to the WZ picture based on the motion vector provided from the motion vector generator 350 using the key picture reconstructed by the key picture decoder 330. And output the supplementary information to the channel code decoder 310. The auxiliary information generator 340 generates side information corresponding to the WZ picture to be reconstructed using interpolation, assuming linear motion between key pictures located before and after the current WZ picture.

The motion vector generator 350 generates a motion vector based on the key pictures and provides the generated motion vector to the auxiliary information generator 340.

In detail, the motion vector generator 350 estimates the motion of a rectangular section or block of a picture.

5 is a view for explaining a process of generating a motion vector according to the present invention.

Referring to FIG. 5, the motion vector generator 350 searches for an area 420 of another picture to find an MⅹN sample area 412 that matches the MⅹN sample block 402 of one picture 400. do. The motion vector generator 350 compares all or part of the MⅹN block 402 of one picture with the possible MⅹN blocks in the search area (generally, the area based on the position of the current block) 420, and the most among them. Find the matching area 412. Subsequently, the motion vector generator 350 generates a motion vector that is a difference value between the position of the current M_N block 402 and the position of the candidate region 412.

Since the size of the macroblock for motion prediction varies in the size of the moving object in the video image, it may be more effective to use various block sizes for motion prediction. In the compression standard of H.264 / Advanced Video Coding (AVC), the block size for motion prediction may be one of 16x16, 16x8, 8x16, 8x8, 8x4, 4x8, and 4x4.

According to the present invention, the motion vector generator 350 determines sizes of a plurality of macroblocks for motion prediction, obtains motion vectors using macroblocks having respective sizes, and selects an appropriate motion vector among the motion vectors. do.

The configuration of the motion vector generator 350 will be described with reference to FIG. 6.

6 is a diagram illustrating a configuration of a motion vector generator 350 according to the present invention. As shown in FIG. 6, the moved vector generator 350 according to the present invention includes a block size determiner 351, A motion prediction unit 352, a motion vector storage unit 353, and a motion vector selection unit 354 are included.

The block size determiner 351 determines the size of the macroblock for motion prediction based on the key picture. The macroblocks may be square and / or rectangular and / or any shape in accordance with the present invention.

In other words, the block size determiner 351 determines a plurality of macroblock sizes. According to an embodiment, the block size determiner 351 may determine large block sizes for motion prediction when the motion between the key features is monotonous and there is no large difference, and there may be fine motion between the key pictures, In many cases, small block sizes can be determined for motion prediction. Large block sizes include block sizes of 16x16, 16x8 and 8x16, and small block sizes may include 8x8, 8x4, 4x8 and 4x4 block sizes. According to another embodiment, the block size determiner 351 may determine the sizes of all macroblocks that can be used for motion prediction for motion prediction. The block size determiner 351 determines the block sizes for motion prediction and provides the determined block sizes to the motion prediction performer 352.

The motion prediction execution unit 352 generates motion vectors by performing motion prediction on each of the sizes of blocks used for motion prediction. In detail, the motion prediction performing unit 352 searches an area of another key picture to find a sample area that matches the sample block of one key picture, finds the block that most matches the difference, and the difference therebetween. Create a motion vector as a value.

The motion prediction performer 352 repeats the process of generating and storing such a motion vector by the number of block sizes.

The motion prediction execution unit 352 stores the generated motion vector in the motion vector storage unit 353.

The motion vector selector 354 measures similarity between blocks by using motion vectors obtained with different block sizes existing in the key picture, and selects the motion vector determined to be the most similar among them. That is, the motion vector selector 354 measures a similarity between the block of the one key picture and the corresponding block of the other key picture among the motion vectors and selects a motion vector having a block size corresponding to the highest similarity. Choose.

Here, the similarity between each block present in the key picture is generally referred to as a sum of absolute difference (SAD) or mean absolute difference (MAD) of blocks present in each key picture. It can be calculated using a method such as the average of the absolute sum of) or SSD (Sum of Square Difference, sum of squares of the difference of each block). In addition, it is also possible to measure the similarity between each block existing in the key picture by applying other appropriate methods in addition to the above-described similarity measuring method.

The motion vector selector 354 provides the selected motion vector to the auxiliary information generator 340.

The operation of the motion vector generator 350 having such a configuration will be described below.

7 is a diagram illustrating an operation of the motion vector generator 350 according to an exemplary embodiment of the present invention.

Referring to FIG. 7, the motion vector generator 350 first determines a plurality of macroblock sizes for motion prediction in step 510.

In this case, the sizes of the macroblocks may be determined according to the characteristics of the image. For example, when there are fine or complex motions between pictures or when there is a lot of motion, small block sizes may be determined for motion prediction. Alternatively, if the motion between features is monotonous and there is no significant difference, large block sizes may be determined for motion prediction. Alternatively, the sizes of all macroblocks that can be used for motion prediction may be determined.

Subsequently, the motion vector generator 350 generates a motion vector by performing motion prediction according to the size of each macro block in step 520. Specifically, the region 420 of the other picture is searched to find the MⅹN sample area 412 that matches the MⅹN sample block 402 of one picture 400. The motion vector generator 350 compares all or part of the MⅹN block 402 of one picture with the possible MⅹN blocks in the search area (generally, the area based on the position of the current block) 420, and the most among them. The matching area, that is, the candidate area 412 is found. Subsequently, the motion vector generator 350 generates a motion vector that is a difference value between the position of the current M_N block 402 and the position of the candidate region 412.

Subsequently, the motion vector generator 350 determines whether the motion vector has been generated for the sizes of all the macroblocks determined in step 530. If a motion vector is generated for all determined macroblock sizes, the motion vector generator 350 measures a similarity between a block of one picture and a block of another picture for each motion vector in step 540. In one embodiment according to the present invention, the block may have a square shape. In addition, as a method of measuring similarity between blocks existing in the key picture, a sum of absolute difference (SAD), mean absolute difference (MAD), or sum of square difference (SSD) may be used. .

However, the present invention is not limited thereto, and the block size may be configured in any shape other than square, and in the method of measuring similarity between blocks existing in the key picture, other arbitrary methods may be selected. . In addition, the present invention can be applied in any case, such as the case of performing the auxiliary information by the interpolation method, the extrapolation method as shown in FIG.

Subsequently, the motion vector generator 350 selects a motion vector corresponding to the highest similarity among the generated motion vectors. That is, a motion vector having the largest similarity between blocks can ensure the best motion compensation.

The motion vector generator 350 outputs the motion vector selected in step 560 to the auxiliary information generator 340.

As described above, in the decoding in the Wyner-Ziv coding technique according to the preferred embodiment of the present invention, a method and apparatus for using selective hash information may be made. Meanwhile, in the above description of the present invention, specific embodiments have been described. Various embodiments may be made without departing from the spirit or scope of the present invention. Accordingly, the scope of the present invention should not be limited by the illustrated embodiments, but should be determined by equivalents of the claims and the claims.

Claims (19)

In the apparatus for decoding a distributed video encoded image using variable block motion prediction for each predetermined coding unit,
A key picture decoder for reconstructing a key picture transmitted from the encoding apparatus;
A motion vector generator configured to determine a size of a block for performing motion prediction using the reconstructed key picture, generate a motion vector, and output the generated motion vector;
An auxiliary information generator configured to generate auxiliary information using the reconstructed key picture and the generated motion vector;
A channel code decoder for estimating a quantized value using the parity bits transmitted from the encoding apparatus and the auxiliary information; And
An image restoring unit for restoring a current picture to be decoded by using the quantized value estimated by the channel code decoding unit and the auxiliary information;
The motion vector generator determines sizes of a plurality of blocks for motion prediction of the auxiliary information, obtains motion vectors between the reconstructed key pictures using blocks having respective sizes, and selects one of the obtained motion vectors. Select the motion vector by
And the plurality of block sizes includes at least one of a size of the coding unit, a size bisected in the horizontal direction, a size bisected in the vertical direction, and a size bisected in the horizontal and vertical directions, respectively. delete The method of claim 1, wherein the motion vector generator
A block size determiner which determines a plurality of block sizes for motion prediction using the reconstructed key picture;
Search for an area of the other key picture of the reconstructed key pictures to find an area that matches the first block of one key picture of the reconstructed key pictures, and find the second block that most matches the A motion prediction performing unit generating a motion vector which is a difference value between a first block and the second block for each of the plurality of block sizes;
A motion vector storage unit for storing motion vectors generated by the motion prediction execution unit; And
A motion vector selector configured to measure a similarity between the first block of the one key picture and the second block of the other key picture among the motion vectors, and select a motion vector having a block size corresponding to the highest similarity; Decoding apparatus for distributed video encoded image. The method of claim 3, wherein the motion vector selector is a sum of absolute difference between the first block and the second block, and an absolute sum of absolute differences between the differences between the blocks. And a similarity measurement using any one of a sum of square difference (SSD) and a sum of square differences (SSD). delete The method of claim 1, wherein the motion vector generator is further configured to estimate the motion by one of a size bisected by the coding unit in the horizontal direction and a size bisected in the vertical direction when the motion is monotonous and there is no significant difference between the reconstructed key features. The apparatus for decoding distributed video encoded images determined for the sake of brevity. The apparatus of claim 6, wherein the size determined for the motion prediction comprises at least one of block sizes of 16 × 8 and 8 × 16. The distributed video of claim 1, wherein the motion vector generator determines fine motions between the reconstructed key pictures or the motion units when the motion vector generation is divided into two horizontal and vertical directions for motion prediction. Decoding apparatus for encoded video. The apparatus of claim 8, wherein the size determined for the motion prediction comprises at least one of block sizes of 8 × 8 and 4 × 4. A method of decoding a distributed video encoded image using variable block motion prediction for each predetermined coding unit,
Restoring a key picture transmitted from the encoding apparatus;
Generating a motion vector by determining a size of a block for performing motion prediction using the reconstructed key picture;
Generating auxiliary information using the reconstructed key picture and the generated motion vector;
Estimating a quantized value using parity bits transmitted from the encoding apparatus and the auxiliary information; And
Restoring a current picture to be decoded using the estimated quantized value and the auxiliary information;
The generating of the motion vector may include determining sizes of a plurality of blocks for motion prediction of the auxiliary information;
Obtaining motion vectors between the reconstructed key pictures using the blocks having respective sizes; And
Selecting a motion vector based on a predetermined criterion among the obtained motion vectors,
The plurality of block sizes include at least one of a size of the coding unit, a size bisected in the horizontal direction, a size bisected in the vertical direction, and a size bisected in the horizontal and vertical directions, respectively. delete 11. The method of claim 10, wherein obtaining the motion vectors
Searching an area of another key picture of the reconstructed key pictures to find an area that matches the first block of one key picture of the reconstructed key pictures and finding a second block that most matches the second block; And
And generating a motion vector, which is a difference between the first block and the second block, for the plurality of block sizes, respectively. The method of claim 10, wherein the selecting of the motion vector
Measuring similarity between the first block of one key picture of the reconstructed key pictures and the second block of the other key picture of the reconstructed key pictures for the respective motion vectors; And
Selecting a motion vector having a block size corresponding to the highest similarity among the motion vectors from which the similarity is measured. 15. The method of claim 13, wherein the similarity is determined by Sum of Absolute Difference (SAD) between the first block and the second block, Mean Absolute Difference (MAD), and Average of the absolute sum of the differences between the blocks. A method of decoding a distributed video encoded image measured using any one of a sum of square difference (SSD). delete The method of claim 10, wherein the determining of the sizes of the plurality of blocks comprises: dividing the coding unit horizontally and bisecting vertically when the motion is monotonous and there is no significant difference between the reconstructed key features. A method of decoding a distributed video encoded image, which is a step of determining one of motion prediction. 17. The method of claim 16, wherein the size determined for motion prediction comprises at least one of block sizes of 16x8 and 8x16. The method of claim 10, wherein the determining of the sizes of the plurality of blocks includes performing motion prediction by dividing the size of the coding unit into two horizontal and vertical directions, respectively, when there is fine motion between the reconstructed key pictures, or when there is a lot of motion. A method of decoding a distributed video encoded image, which is a step of deciding a risk. 19. The method of claim 18, wherein the size determined for motion prediction comprises at least one of block sizes of 8x8 and 4x4.

Images