KR970004926B1 - Selective error correction method for digital image transmission - Google Patents

Abstract

The present invention is related to a selective error correcting method of a digital picture transmission. In an error correcting method where a picture signal is divided into six classes according to a MPEG picture compression method and is transmitted, only data of the four classes, i.e., each header data, a DC coefficient of each block of an intra picture, a MV of a prediction mode macro block of a predictive picture, and a DC coefficient of a macro block of an intra picture are encoded as a RS which is an error correcting code. The error correcting code is converted into a parity signal and transmitted by adding at each slice end.

Description

Selective Error Correction Method for Digital Video Transmission

1 is a schematic diagram illustrating a selective error correction method of an I screen according to the present invention.

2 is a schematic diagram for explaining a selective error correction method of a P screen according to the present invention.

The present invention relates to a selective error correction method of digital image transmission, and in particular, image data compressed by a differential pulse coded modulation / discrete cosine transform (hereinafter referred to as DPCM / DCT) complex coding method. For compensating for bit error or cell loss when transmitting over a broadband band-integrated service digital network (hereinafter referred to as B-ISDN) in an asynchronous transfer mode (hereinafter referred to as ATM). Correction code compensates for transmission error with only a small amount of data by reducing the amount of data added by using error correction code for only some data in the compressed bit string where the effect of transmission error is large, such as video data control signal or DPCM data. And selective error correction method of digital HDTV video transmission to improve the restored picture quality. One will.

Currently, high definition televisions (hereinafter referred to as HDTVs), data storage media, multimedia and digital TVs compress and transmit data in a predetermined manner to transmit a large amount of data. And compression methods are not standardized, but many methods have been studied due to high utility such as HDTV, and standardization is steadily progressing.

ATM is a method of transmitting a video signal through B-ISDN, which is standardized by International Telecommucation Union (ITU) -Telecommunication Standardization (TS). The data compression method used for video communication through the ATM transmission network includes a variable bit rate (VBR) encoding method and a fixed bit rate (hereinafter referred to as CBR) encoding method. The VBR encoding method has been mainly studied due to the characteristics of the CBR encoding method. The CBR encoding method is mainly focused on the expert group of SG15 of ITU-TS.

In the expert group for the standardization of the video compression method used for ATM video communication under the ITU-TS SGI15, the standardization of the video compression method is also carried out together with another expert group, moving pecture experts group (hereinafter referred to as MPEG). Doing. In addition, the digital HDTV system, which is being enacted by the US Federal Communications Commission, also uses video compression technology similar to that of MPEG.

Therefore, the most representative video compression method developed to date can be called MPEG video compression method.

There is always noise in all transmission paths, which causes bit errors in digital transmission networks. In addition, in an ATM transmission network, a cell may be damaged due to channel damage and network congestion in addition to a bit error. However, if a compressed video signal is very sensitive to a transmission error and a cell is lost in the ATM transmission process, the decoder that recovers the video loses synchronization and the restored video is greatly damaged.

As a method of restoring an image error as described above, an error recovery method for recovering lost data using an error gain method and an error correcting code is conventionally used.

However, the conventional error gain method can obtain some effects by using an error gain method suitable for image encoding, but when the transmission error occurrence rate increases, the effect drops rapidly.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to improve the quality of reconstructed video by applying an error correcting code only to some data that is an important factor for bit string recovery in video signal data compressed by MPEG method. In addition, the present invention provides a selective error correction method for digital video transmission that can reduce the transmission speed by reducing the burden caused by an increase in the amount of data due to an error correction code.

A feature of the selective error correction method of digital video transmission according to the present invention for achieving the above object is a cyclic symbol for data that is absolutely necessary for reconstruction such as header data in each cell of encoded video data according to the MPEG video compression method. It is characterized by reducing the amount of additional data by encoding error information.

Hereinafter, a selective error correction method of digital image transmission according to the present invention will be described in detail with reference to the accompanying drawings.

First, the error correcting code according to the present invention uses Reed-Solomon (hereinafter, referred to as RS) code, which is a cyclic symbol error correcting code. The RS code is used in the present invention because it is the most efficient error correcting code that can be used in an ATM transmission method because it is very useful for burst errors because the RS code can correct an error in symbol units.

In the RS code (N, K, D), N denotes the number of output symbols of the encoder, K denotes the number of input symbols of the encoder, and d denotes the maximum modifiable data amount of the code. If e is an error symbol and E is an erased symbol, it appears as 2e + Ed.

However, the ATM application layer examines a sequence number assigned to each ATM cell to detect lost cells and cells in which an error has occurred.

Therefore, since the location of the lost cell and the error cell can be clearly known in the ATM transmission network, it is replaced by the symbol in the lost cell. Therefore, if the RS code is used, the cell is twice as large as the bit error occurs. Corrective effect can be obtained.

In addition, in the present invention, after determining the type of the error correcting code as the RS code, it is important to select data to which the error correcting code is to be applied and data not to apply the error correction code to only a part of the bit stream compressed in the ATM video signal.

Here, an example of the most representative MPEG2 image compression method among DPCM / DCT complex coding methods will be described.

In the MPEG2 video compression method, when the video is encoded by compressing the input images, the compressed bit stream has the start and end of each layer in addition to the DCT coefficients, which are divided into the most basic data, the motion vector and the average brightness DC coefficient, and the AC coefficient representing the image. A large variety of data, such as data representing various parameters and other parameters, is divided into six levels of layer compression.

That is, the six steps

One). Sequence start code, sequence end code, group of picture (hereinafter referred to as GOP) start code, GOP end code, and picture start code (pictrre) Header data such as atart code, pictrre end code, slice start code, slice end code, and the like.

Then, as the image layer, in the three encoded types of images, intra (hereinafter referred to as I) image, unidirectional predictive (hereinafter referred to as P) image and bidirectional predictive (hereinafter referred to as B) image

2). I DC and AC coefficients of the DCT coefficients of the angular block of the image.

3). A moving vector (hereinafter referred to as MV) of a prediction mode MB (macroblock) of P and B images.

4). MB block (hereinafter referred to as MBA) increment.

5). AC coefficient among DCT coefficients of each block.

6). Divide by the DC and AC coefficients of the I mode macroblock.

In this case, all data generated in the layer higher than MB are separated into header data.

The data of each layer is described in detail to select data to be error-corrected encoded from the data of each layer as follows.

First, since various header data are data representing the encoding layer of MPEG, if an error occurs in the header data, the GOP or the display sequence cannot be largely restored in one smaller slice. Accordingly, the header data are separately RS-coded to restore the RS-coded data into the ATM cell when the ATM cell is damaged. For example, in a sequence layer, an aspect ratio, an aspect ratio, an image rate, a bit rate, and the like of an image are RS encoded, and an extension code or user data is not RS encoded.

Then, the above data is put in an ATM cell, and one cell is added to put parity data. In this case, if the symbol size is increased or the number of parity data is increased to improve error correction capability, the loss of cells can be recovered. For example, all header data can be put in about four ATM cells using an RS code whose symbol size is 8 bits (31, 21) in the header data.

In addition, the error correction coding for data below the image layer is as follows.

In the video layer, RS encoding is performed on an image encoding type, virtual buffer data, and motion prediction parameters.

MB is necessarily coded in I mode, and among the DCT coefficients of each block, AC is scalar quantized and then represented by a variable length code (hereinafter referred to as VLC), and DC is encoded by the DPCM method.

If a transmission error occurs, synchronization is lost in the VLC decoding process due to the characteristics of the variable length code.

At this time, for the VLC resynchronization, the start point of the bit string of each MB is transmitted in the data transmission process of the encoder or a synchronization signal having a fixed bit is provided for each MB so that synchronization is performed in units of macro blocks. Even if the variable length code is decoded after synchronization, the DC value is not correctly restored because errors are accumulated in the DPCM decoding process because of the lost information. A similar phenomenon occurs because the motion vector and mba of the prediction mode MB are also DPCM coded in the P or B image. On the other hand, all DCT coefficients of AC and predictive MB among the DCT coefficients of the I mode MB can be correctly restored if the variable length information is recovered. Since mba can be restored after the brake is made in the VLC decoding process, there is no problem.

If cell loss occurs when transmitting an ATM cell containing an MPEG video signal bit string, an error occurring in the VLC decoding process is not limited to the corresponding cell but continues to affect a plurality of cells transmitted thereafter.

In the method of the present invention, the error correction code is applied only to some important information in order to minimize the accumulation of errors in the decoding process while reducing the amount of additional information required for error correction encoding.

Information applying the error correction code in the MPEG video compression method is as follows.

One). Various header information.

2). I DC coefficient of each block.

3). P picture prediction mode MB MV.

4). DC of I mode MB.

In the B image, the error correction code was not used because the MV of MB included in one cell is too inefficient to use the error correction code compared to the other images.

FIG. 1 is a schematic diagram illustrating an RS encoding process of an I image according to the present invention. RS encoding is performed on a slice in an I image by a basic unit, and RS coding is applied to DC coefficients of each block.

That is, RS coding is performed by collecting only DPCM prediction errors of DCs of all MBs included in the slice. In this case, when the sysyematic code is used, if the original information is not changed and only a parity code is added, the parity code is added at the end of each slice.

2 is a schematic diagram for explaining an RS encoding process of a P picture according to the present invention, and an application principle for RS coding an MV in a P picture is the same as that of encoding a DC of an I picture.

On the other hand, if I-mode MB occurs in P, then DC coefficients of the corresponding block are encoded together with the motion vectors of all the MBs of the prediction. Using this method, since the amount of data added in the RS encoding process per unit cell is very small, error correction encoding can be performed very efficiently.

Then, data which is not recovered even by the above correction method by error correction coding forms a screen by an error concealment method.

That is, after detecting the synchronization of the VLC code again in the bit stream where the error occurs, the information contained in the lost cell should be recovered. In this case, the information recoverable using the RS-encoded information is the number of DCs of the I image, the MV of the P image, and the number of DCs of the I mode of the P image as described above.

Therefore, the lost MB conceals errors by properly using the information of the surrounding MB. In order to increase the effect of the error concealment, characteristics of an image encoding and an algorithm are used.

First, when the transmission error is restored, the following error is concealed.

The error concealment method in the I picture is 4: 1: 1 of the color format of the input picture. Since there are 6 blocks in one MB, the black and white signal is referred to as the two blocks at the top. Replace with two blocks at the bottom of the MB above it. The color signals Cb and Cr are compensated by using a portion corresponding to the Y signal. In this case, since DC information has already been restored, only data obtained by subtracting DC values of the corresponding blocks is used.

In addition, the error concealment method in the P picture uses only the MV in the recovered MB to compensate only the motion from the previous picture and does not compensate for the motion compensation prediction error signals.

The error concealment method in the B picture then compensates from the previous picture using the MV of the surrounding MB.

As described above, the selective error correction method of the digital video signal according to the present invention includes various header information, DC coefficients of each I video block, and P video prediction mode MB among six video signals of the MPEG video compression method. Since only RS data of 4 layers, such as MV and I mode MB of DC coefficient, is RS-encoded, the error correction code is converted into a parity signal, and transmitted at the end of each slice. The reduced information amount, which is reduced by more than 90%, can be used in the video encoding process, thereby improving the image quality of the restored image. In addition, the selective error correction coding method according to the present invention is useful for a system requiring a low-delay coding mode such as HDTV since the PSNR of the reconstructed picture is gradually reduced and the excellent picture can be obtained without the B picture even if the CLR is increased. There is this.

Claims (1)

An error correction method of digital video transmission in which video signals are separated and transmitted by six layers according to the MPEG video compression method, comprising: various header information among the six layers of video signals, DC coefficients of each block of I video, and P video RS coding, which is only error correction code, is performed on the 4th layer data such as the MV of the prediction mode macroblock and the DC coefficient of the I mode macroblock, and the error correction code is converted into a parity signal and added to the end of each slice for transmission. Error correction method.