KR20030033123A - Error concealment device using samples of adjacent pixel and method thereof - Google Patents

Abstract

PURPOSE: An apparatus and a method for concealing an error using samples of peripheral pixels are provided to reduce time required for concealing the error by efficiently processing the error generated in an MPEG-2 bit string by sampling peripheral pixels. CONSTITUTION: An error detecting part detects whether an error is generated in a current frame. The mean absolute value error of a boundary pixel is calculated in a macro block of a former frame located in the same position as a macro block of the current frame(100) in which errors are generated. The mean absolute value error is compared with a preliminarily set threshold(101). If the mean absolute value error is smaller than the threshold, the macro block is copied from the former frame(102). If the mean absolute value error is larger than the threshold, peripheral pixels of the macro block are sampled at a certain rate(103). The mean absolute value error is calculated by setting a whole area as a search range corresponding to the sampled peripheral pixels(104). Another macro block having the smallest mean absolute value error is copied from the former frame corresponding to the calculated mean absolute value error(105).

Description

Error concealment device using samples of adjacent pixel and method

The present invention relates to an error concealment apparatus and method using samples of surrounding pixels, and more particularly, to samples of corrupted macroblock surrounding pixels in a previous frame that has already been decoded and reconstructed when the macroblock information of the current frame is damaged together with the motion vector information. And an error concealment apparatus and method using samples of neighboring pixels that find and recover an error of a macroblock having minimum mean absolute error (MAE).

Recently, with the rapid digitization of images, the use of digital TV is rapidly increasing due to the development from the existing analog TV image. Digital video requires a lot of memory and bandwidth when storing and transmitting information due to the large amount of representational information. To solve this problem, a method for reducing transmission bit rate or storage capacity of many compression coding technologies has been developed.

In this way, video compression coding technology using MPEG technology is generally used. Among the bit streams of MPEG-2, variable cosine transform (DCT), quantized coefficients, and motion vector information found in macroblock units are variably encoded. Therefore, when an error occurs in the macroblock, the influence of the error is not limited to one pixel, but spreads in macroblock or slice units so that information is lost.

In addition, MPEG-2 has four layers in the MPEG-2 hierarchy, namely, a video sequence layer, a group of pictures layer, and a picture layer to solve a synchronization problem in a bit stream. There is a fixed length start code in the Layer and the Slice Layer. Therefore, the propagation effect of the transmission error will continue until the decoder finds the next start code, resulting in large degradation of the images. In order to solve this phenomenon, an error concealment method using neighboring pixels of a damaged macroblock has been studied.

On the other hand, a transmission error generated during transmission of a digital video can be largely classified into a random error and a burst error. First, the random error affects only one codeword in the case of a fixed length coding scheme, which is generally decodable. However, in the case of the variable length coding scheme, the random error shifts the synchronization of the coded information, so that subsequent codewords cannot be decoded. In addition, the continuous error appears in the form of a continuous error due to packet loss or physical defect of the storage device in the packet network, which has a more serious effect on the data transmission system than a random error. In addition, most recent still image / video compression techniques adopt a variable length coding scheme. When a random error occurs in such a system, continuous data is distorted as if a continuous error occurs.

That is, if an error occurs during data transmission and the information of the motion vector is also damaged, the image is not decoded and the output screen is broken. Therefore, an error concealment technique for motion compensation of an errored block is very important on the decoder side. At this time, there are various ways to restore the motion vector. First, there is a method of taking an average of surrounding motion vectors, or a method of taking a median. However, the limitation of the method of using the surrounding motion vector is that the motion of the damaged macroblock has a high correlation with the motion of the neighboring block, and if the motion only occurs or the motion does not occur only in the damaged macroblock, the restored image is moved. The problem is that the object disappears or causes movement.

Conventional techniques for concealing the above-described errors are described in "Recovery of lost or erroneously received motion vectors, Proceedings ICASSP'93 pp v417-v420". However, this conventional error concealment method uses a large amount of computation because it uses all the pixels around the damaged block.

The present invention has been made to solve the above problems, and when a region for reconstructing a damaged macroblock is found, a macro whose MAE is minimized within a constant search range of a previous frame by extracting a sample of boundary pixels around the damaged macroblock. An error concealment apparatus and method using samples of surrounding pixels to find a block and reconstruct a damaged macroblock.

1 is a block diagram of an error concealment apparatus using samples of surrounding pixels according to the present invention;

2A and 2B show damaged macroblocks and neighboring macroblocks.

3A and 3B illustrate an embodiment in which a peripheral pixel is sampled 2: 1 in the present invention.

4A and 4B illustrate an embodiment in which a peripheral pixel is 4: 1 sampled in the present invention.

5 is a flow chart of an error concealment method using samples of surrounding pixels in accordance with the present invention.

6 to 8 are graphs showing the error concealment time of the present invention.

<Description of Symbols for Main Parts of Drawings>

10: error detection unit 20: error concealment control unit

30: decoding unit

In order to achieve the above object, an error concealment apparatus using samples of neighboring pixels of the present invention includes an error detection unit that receives an encoded bitstream and detects whether an error occurs, and an error in inputting an error generation signal from the error detection unit. Sample the samples of the surrounding pixels in the previous frame of the generated macroblock, copy the macroblock having the smallest mean absolute value error in a certain search range, and input an error concealed controller and an error concealed image from the error concealment controller. And a decoding unit for decoding and outputting the received signal.

Also, an error concealment method using samples of neighboring pixels of the present invention includes a first step of detecting whether an error has occurred in a current frame, and a macro of a previous frame at the same position as a macroblock of the current frame in which the error has occurred. A second step of calculating an average absolute value error of the boundary pixel in the block; a third step of comparing the average absolute value error calculated in the second step with a preset threshold; If smaller than the value, the macroblock of the same position is copied from the previous frame, and if large, the fourth region of sampling the peripheral pixels of the macroblock at a predetermined rate and the entire region corresponding to the peripheral pixels sampled in the fourth stage It has the smallest average absolute value error corresponding to the average absolute value error calculated in the fifth and fifth steps of calculating the average absolute value error. Is characterized in that it comprises a sixth step of copying the macroblock in the previous frame.

Hereinafter, with reference to the accompanying drawings will be described in detail an embodiment of the present invention.

In the present invention, by setting the search range using only the sample pixels around the damaged macroblock, the calculation amount can be drastically reduced compared with the conventional method using all the boundary pixels around the error block. This technique of the present invention can be applied as an efficient error concealment algorithm to a DTV video decoder using MPEG-2.

1 is a block diagram of an error concealment apparatus using a motion vector of a neighboring macroblock according to the present invention.

Referring to FIG. 1, the present invention relates to an error detection unit 10 that receives an encoded bitstream and detects whether an error has occurred, and to a previous frame of a macroblock in which an error occurs when an error occurrence signal is input from the error detection unit 10. An error concealment control unit 20 for sampling the samples of the surrounding pixels at 2: 1 or 4: 1 to copy a macroblock having the smallest Mean Absolute error (MAE) in a constant search range to conceal an error; and The error concealment control unit 20 is configured to receive an error concealed image from the decoding unit 30 for decoding and outputting.

In the present invention, all random errors and continuous errors are classified into transmission error ranges.

2A and 2B show corrupted macroblocks of the current frame and pixels around the macroblock of the previous frame at the same position as the current frame.

Here, Ci and Pi mean peripheral pixel values of the current frame and the previous frame block, respectively.

Referring to the operation process of the present invention, first, the error detection unit 10 of FIG. 1 detects whether there is a damaged macroblock as shown in FIG. 2 and outputs an error detection result to the error concealment control unit 20.

Here, the equation for calculating the average absolute value error for the full range search in the error concealment control unit 20 is as follows.

Mean absolute error MAE = │Ci-Pi│ ....... (1)

Equation (1) calculates MAE, where N is the number of neighboring pixels. In the case of a 16 × 16 macroblock, the number N of neighboring pixels is 68. If the MAE value is smaller than the set threshold, it is determined that there is no motion and the macroblock of the previous frame at the same position as the current frame is copied as it is. Otherwise, it is determined that there is motion and search for the full range. Find the block that minimizes MAE between neighboring pixels.

Here, the reason why the threshold is set is that the MAE of the previous frame at the same position as the current frame will be sufficiently small if there is almost no movement such as the background while the screen is fixed. Therefore, this threshold can be used to obtain effective results in terms of both computation and execution time without unnecessary search. The method of finding the macroblock that minimizes the MAE of the neighboring pixels through the full range search is the same as that of finding the motion vector within the search range when the motion estimation is performed, and thus has the highest similarity to the motion vector of the block.

At this time, since the correlation between adjacent pixels in the surrounding pixels is very high, the calculation amount can be significantly reduced by sampling the neighboring pixels to find the minimum macroblock position by calculating the MAE. Objective image quality can be obtained.

3A and 3B show an example in which the error concealment control unit 20 samples the neighboring pixels of the macroblock in which an error occurs at 2: 1.

MAE = │C2i-P2i│ ....... (2)

In Eq. (2), the number of neighboring pixels N is 34 for 16 × 16 blocks.

4A and 4B show an example in which the error concealment control unit 20 samples 4: 1 of the pixels around the macroblock in which the error occurs.

MAE = │C4i-P4i│ ....... (3)

In Eq. (3), the number of neighboring pixels N is 17 for 16 × 16 blocks.

Blocks indicated by dark portions in FIGS. 3 and 4 are sampling values of the number N of neighboring pixels used for the calculation of the MAE. That is, FIG. 3 samples one of two values, and FIG. 4 samples one of four values.

5 is a flowchart illustrating an error concealment process using samples of neighboring pixels in the error concealment control unit 20.

First, when an error occurs and the macroblock is damaged, the MAE of the neighboring pixel is calculated in the macroblock of the previous frame at the same position as the neighboring pixel of the damaged macroblock of the current frame (step 100). If the value of the MAE is smaller than the predetermined threshold Th, it is determined that there is no motion (the motion vector is (0,0)) and the macroblock at the same position from the previous frame is copied to the current frame (step 102). If the value of the calculated MAE is larger than the predetermined threshold Th, the peripheral pixels of the macroblock are sampled at 2: 1 or 4: 1 (step 103). Next, the previous sampled at 2: 1 or 4: 1. The entire area in the frame is set to the search range [-R, + R-1]. (Step 104) Here, the range [-R, + R-1] of the search region of the present invention is set to [-15, 16]. Set it. Next, the MAE of the entire area is calculated within the sampled search range and the macro block having the smallest MAE is copied to restore the damaged macroblock of the current frame (step 105).

6 to 8 show error concealment time of an error concealment method using sampling of the surrounding pixels of the present invention.

FIG. 6 is a graph showing error concealment time according to a football sequence, FIG. 7 is a graph showing error concealment time according to a flower garden sequence, and FIG. 8 is a graph showing error concealment time according to a table tennis sequence.

6 to 8, it can be seen that the present invention reduces the error concealment time when sampled at 2: 1 and 4: 1, respectively, compared to the method using all the surrounding pixels. Through these experimental results, it can be seen that the method of concealment through sampling proposed by the present invention is a method for minimizing error concealment time.

As described above, the present invention provides an effect of reducing an error concealment execution time by efficiently processing an error occurring in an MPEG-2 bit string through sampling of neighboring pixels.

Claims (4)

An error detector which receives an encoded bitstream and detects whether an error occurs; An error concealment control unit for hiding an error by sampling a sample of neighboring pixels in a previous frame of the macroblock in which an error occurs when an error generation signal is input from the error detection unit, copying a macroblock having the smallest mean absolute value error in a predetermined search range. ; And An error concealment apparatus using samples of neighboring pixels, characterized in that it comprises a decoding unit for receiving an image concealed error from the error concealment control unit to decode the image. The method of claim 1, wherein the error concealment control unit MAE = │Ci-Pi│ The error concealment apparatus using samples of neighboring pixels, wherein the average absolute value error is obtained using the following equation (N is the number of neighboring pixels, Ci and Pi are the neighboring pixel values of the current frame and the previous frame block, respectively). Detecting whether an error has occurred in the current frame; A second step of calculating an average absolute value error of a boundary pixel in a macroblock of a previous frame at the same position as the macroblock of the current frame in which the error occurs; A third step of comparing the average absolute value error calculated in the second step with a preset threshold value; A fourth step of copying a macroblock at the same position from a previous frame when the average absolute value error is smaller than a threshold value in the third step, and sampling peripheral pixels of the macroblock at a predetermined ratio when the average absolute value error is smaller than a threshold value; A fifth step of calculating an average absolute value error by setting the entire area as a search range corresponding to the peripheral pixels sampled in the fourth step; And And a sixth step of copying a macroblock having the smallest average absolute value error in a previous frame in response to the average absolute value error calculated in the fifth step. The method of claim 3, wherein the fifth step MAE = │Ci-Pi│ The method of error concealment using samples of neighboring pixels, wherein the average absolute value error is obtained using the equation (N is the number of neighboring pixels, Ci and Pi are the neighboring pixel values of the current frame and the previous frame block, respectively).