KR100375345B1 - Video coding method and apparatus thereof for error resilient mode - Google Patents

Abstract

A video coding method and coding apparatus are disclosed. This video coding method includes a macroblock including header data, motion vector data, and discrete cosine transform (DCT) data, respectively, in order to enable less sensitive communication in an error-prone environment such as a wireless communication channel. Discrete cosine transform in which the header data is grouped into a header data area (HDP), the motion vector data area (MVDP) in which the motion vector data is grouped, and the discrete cosine transform data are grouped. After data partitioning into a data region (DDP: DCT Data Part), regions selected according to a predetermined priority among the divided regions are inversely variable-coded and the remaining regions are variable-length coded.

Description

Video coding method and apparatus for error-tolerant mode TECHNICAL FIELD

The present invention relates to a video coding method and an apparatus thereof, and more particularly, to a video coding method and an apparatus for an error-tolerant mode. The invention is also based on U.S. Provisional Application No. 60 / 067,013, entitled "Video Codec Method in Error Resilient mode," filed by the applicant of the present application. The error-tolerant mode refers to a data transmission mode that can recover even a certain amount of error in the transmitted bitstream. In this case, the greater the degree of error recovery, the higher the tolerance.

Because video codecs are primarily concerned with source coding rather than channel coding, they typically do not have the ability to cope with errors when transmitting bitstreams over channels that are prone to error. This is because the video codec has no judgment element for determining whether to restore or discard an errored portion of the bitstream. That is, when an error occurs, it is difficult to determine whether there is a bit loss in only one part of one frame or an error in the entire frame, so that the bitstream of the entire frame in which the error occurs is discarded and the bitstream for the next frame is lost. There is a problem to find the starting point of. In order to solve this problem, for example, in the conventional coding method, a picture start code (PSC) is provided, and if an error is received after the picture start code, the following part is ignored and the next picture start code ( PSC) can also be used to find them. In addition, a GOB Start Code (GBSC) indicating the start of a group of blocks (GOB) is set. If an error is received after this information, the next block group is ignored and only the part below it is ignored. Finding is reducing the area discarded.

1 shows an example of a video data packet generated by a coding method for an error-tolerant mode according to the prior art. As shown in FIG. 1, it can be seen that the coding method for the conventional error-tolerant mode separates coding into motion data and texture data. In more detail, the motion data includes a macroblock identification region (COD) indicating coding, a macroblock pattern chromaticity region (MCBPC) indicating a chromaticity type of each macroblock, and a motion vector region. The texture data includes a CBPY region (Coded Block Pattern luminance (Y)), a DQUANT region (Data Quantization), and a DCT region (Discrete Cosine Transform). Such motion data and texture data are separated by a motion marker (MM). Herein, the discrete cosine transform data is RVLC (Reversible Variable Length Coding). In addition, a byte alignment process is performed in the process of inserting a Resync Marker (RM).

In decoding, if a resynchronization marker RM is found in the bitstream, the interval until the next resynchronization marker RM is found is considered as one packet. Meanwhile, since only motion vectors predicted based on previous data exist in the Motion Vector Data Part (MVDP) consisting of motion vectors, the motion vectors can be used only when there are previously decoded motion vectors. Do. Therefore, if an error is found in the motion vector data area (MVDP) in the bitstream, there is a problem that there is a lot of information loss until the resynchronization marker RM where the next packet starts is found, that is, the entire packet must be ignored. .

An object of the present invention is to provide a video coding method with low information loss and high error tolerance.

Another object of the present invention is to provide a computer readable recording medium for executing the video coding method.

Another object of the present invention is to provide a video coding apparatus for implementing the above method.

1 is a diagram illustrating an example of a video data packet generated by a conventional video coding apparatus.

2 is a flowchart illustrating main steps of a video coding method according to an embodiment of the present invention.

3 is a block diagram illustrating a video coding apparatus according to an embodiment of the present invention.

4 is a diagram illustrating an example of a video data packet generated by a video coding method according to an embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

20 splitting step, 22 variable coding step,

24 ... Potency variation coding step, 26 ... Marker insertion step.

In order to achieve the above object, a video coding method according to the present invention includes a header block (HDP) in which header data is grouped into macroblocks including header data, motion vector data, and discrete cosine transform (DCT) data. Data partitioning into a motion vector data region (MVDP) in which the motion vector data is grouped, and a discrete cosine transform data region (DDP: DCT Data Part) in which the discrete cosine transform data is grouped. Doing; And variable-length coding or inverse-variable coding the divided regions.

To achieve the above object, a computer-readable recording medium for executing a video coding method according to the present invention includes macroblocks each including header data, motion vector data, and discrete cosine transform (DCT) data. A header data area (HDP: Header Data Part), a motion vector data area (MVDP) in which the motion vector data is grouped, and a discrete cosine transform data area (DDP: DCT Data) in which the discrete cosine transform data is grouped Program code for data partitioning into data partitions; And program code for inversely variable coding the regions selected according to a predetermined priority among the divided regions and variable length coding the remaining regions.

In accordance with another aspect of the present invention, a video coding apparatus according to the present invention includes a header block (HDP) in which headers are grouped into macroblocks including header data, motion vector data, and discrete cosine transform (DCT) data. Data Part), a motion vector data area (MVDP) in which the motion vector data is grouped, and a discrete cosine transform data area (DDP: DCT Data Part) in which the discrete cosine transform data is grouped (Data) means for partitioning; And means for variable-length coding or inverse-variable coding the divided regions.

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

2 shows main steps of a video coding method according to an embodiment of the present invention. Referring to FIG. 2, the video coding method according to the present invention includes a segmentation step 20, a variable length coding step 22, an inverse variance coding step 24 for selected regions, and a marker insertion step 26. It is done by

In the partitioning step 20, data partitioning for a plurality of macroblocks is performed. Data partitioning means rearranging data constituting the macroblock into a new combination. More specifically, in the present embodiment, the header data area (HDP: Header Data Part) is configured by grouping header data included in each macroblock, and grouping motion vector data and discrete cosine transform data included in each macroblock, respectively. To form a motion vector data area (MVDP) and a discrete cosine transform data area (DDP).

The header data includes information about the coded state of the current macroblock. That is, the header data indicates whether the current macroblock is an intra macroblock in which the contents of the current frame are coded as it is or an inter macroblock in which the difference with the previous frame is coded. Very serious information loss occurs. As the header data is of such high importance, it is necessary to restore the header data first. Thus, for example, the macroblock identification (COD) and the macroblock pattern chromaticity (MCBPC) indicating the coding status in the header data may be combined and divided into one new syntax (COD + MCBPC). The reason why it is preferable to combine and divide in this way is as follows. In the current H.263 standard, one bit is allocated to COD as header data, and the MCBPC is variable-length coded. It is difficult to identify whether an error has occurred because a COD can be assigned one bit so that both 0 and 1 can exist. In addition, when an error occurs in the MCBPC using the variable length code, it is difficult to detect the error because the bit string in which the error occurs is more likely to exist in the variable length code table than the inverse variable code table. Therefore, according to an embodiment of the video coding method according to the present invention, it is possible to easily identify whether an error has occurred by combining the COD and the MCBPC at the time of splitting, and performing the reverse modulation coding of the combined region. For example, the reverse variation codeword 0110 can be identified as an error even if an error such as 0101 or 1001 occurs, thereby reducing the possibility of using information including the error. Moreover, backward recovery is also possible since it is coded with inverse variance coding. However, if necessary, the header data area HDP may be configured through a method of coding each part without using a method of combining and coding each part.

In addition, since the dividing step 20 is performed in units of macroblocks, the first macroblock address (MBA) indicating the number of the first macroblock included in each video data packet is inserted to be referred to during decoding. Do it. In the dividing step 20, the final motion vector value (LMVV) is inserted after the motion vector data area MVDP. Here, the final motion vector value means absolute motion vector information, not a value predicted based on the previous motion vector value. Further, in the dividing step 20, a packet number (PN) indicating a serial number of each video data packet is inserted as packet header information.

In the variable length coding step 22, variable length coding is performed on at least some of the header data area HDP, the motion vector data area MVDP, and the discrete cosine transform data area DDP. At this time, the packet number PN and the final motion vector value LMVV are also variable length coded.

In the reverse variable coding step 24, the reverse variable coding is performed on regions selected according to a predetermined priority required for reconstruction. In other words, the reverse modulation coding is performed on important information that may be lost when decoding the entire video data packet or macroblock. The reason why priority is needed is described in detail as follows. As described above, the header data indicates whether the current macroblock is an intra macroblock in which the content of the current frame is coded as it is or an inter macroblock in which the difference of the previous frame is coded. If an error occurs in the header data of the inter macroblock, motion vector data or discrete cosine transform data cannot be used. If an error occurs in motion vector data, discrete cosine transform data cannot be used. Moreover, most macroblocks are composed of inter macroblocks to increase compression efficiency. Therefore, the importance is high in order of header data, motion vector data, and discrete cosine transform data. In this way, since there is a priority according to the importance in the macroblock, it is preferable that an error is also restored according to the priority. The reason for selectively inverting the data by giving priority to all data without performing the inverse modulation coding is to consider the limited channel capacity because the number of bits is increased when the variable length coding is performed. Although not shown in FIG. 2, the video coding method according to the present invention detects characteristics of a channel by receiving information on characteristics of a channel such as a transmission capacity of a channel, an error situation, and congestion through a back channel. Next, if the channel characteristic permits, reverse-varying coding (RVLC) down to a lower priority region, such as a discrete cosine transform data region (DDP: DCT Data Part), and the priority is not Variable length coding (VLC) is performed in order of low region. At this time, information guiding the inversely variable length coded region and the variable length coded region may be inserted as packet header information. As such, according to the present embodiment, the error tolerance may be increased as much as possible according to the channel characteristics. Alternatively, when the channel characteristic allows, additional information is inserted into a region having a lower priority such as a discrete cosine transform data region (DDP: DCT Data Part) so that it can be referred to when decoding. Hereinafter, the use of such additional information will be described by way of example. First, it is possible to transmit the information of the header data area once more. This is for the case where an error occurs in the header data area of high importance and an error may not occur in the motion vector data area and the discrete cosine transform data area. In this way, by placing the header data area immediately after the discrete cosine transformed data area, it is possible to restore the video data packet without discarding it. Second, it is possible to transmit the CRC as additional information in the discrete cosine transform data region. For example, the discrete cosine transform data uses a fixed length code (FLC) of 8 bits for a DC value and a variable length code (VLC) for an AC value. An 8 bit fixed length code is relatively difficult to detect errors. Resilience can be improved by adding a CRC to this 8-bit fixed length code to help detect errors. In addition to the 8 bits of the fixed length code, other information using the fixed length code, for example, a quantization parity (QP) indicating a quantization value may be used to generate the CRC. In this way, CRC is added to bits of the fixed-length code in the discrete cosine transform data region, thereby improving error recovery performance. As a result, the above-described additional information of the two cases is for restoring an errored portion in the corresponding video data packet or for helping to detect an error more easily in a portion where error detection is difficult. In addition, in the video coding method according to an embodiment of the present invention, error detection is more effectively performed by using different inverse modulation coding tables according to channel characteristics when inverse modulation coding.

In the marker insertion step 26, a resync marker (RM) for distinguishing each video data packet and a header marker (HM) for distinguishing the header data region and the motion vector data region are inserted. In addition, a motion marker (MM) is inserted to distinguish the motion vector data area (MVDP) and the discrete cosine transform data area (DDP). The header marker HM is made using a codeword that is not used for coding the header data area HDP, so that the header marker HM can be easily identified during decoding.

3 shows a video coding apparatus according to an embodiment of the present invention for implementing the method. Referring to FIG. 3, the video coding apparatus according to the present invention titers partitioning means 30, variable length coding means 32, and regions selected according to a predetermined priority required for reconstruction among the regions to be variable length coded. Inverse modulation coding means 34 for disguising coding are provided. In addition, the video coding apparatus includes channel identification means 36 for identifying characteristics of the channel by receiving information on characteristics of the channel, such as channel capacity, error conditions, and congestion, for example, through a back channel (not shown). And marker inserting means 38 for inserting a marker into variable length coding or inversely variable coding bit regions.

The dividing means 30 receives video data and divides the data into a header data area HDP, a motion vector data area MVDP, and a discrete cosine transform (DCT) data area DDP as described with reference to FIG. Perform data partitioning. In addition, the dividing means 30 combines a plurality of predetermined portions having high importance in restoration such as, for example, macroblock identification (COD) and macroblock pattern chromaticity (MCBPC) in the header data area HDP. It can be configured with new syntax (COD + MCBPC).

The variable length coding means 32 performs variable length coding on the divided regions, and the reverse length coding means 34 restores, for example, a portion in which the entire video data packet or a macroblock may be lost if it is lost during decoding. Inversely, variable-variable coding with better error resilience as compared to variable-length coding is performed on regions selected according to a predetermined priority required for the. However, since the number of bits is increased when the variable length coding is performed, it is preferable to perform the reverse length coding only on the bit regions selected with priority in consideration of the limited channel capacity.

To this end, if it is determined that the channel characteristic is allowed, the channel characteristic identification means 36 performs reverse variation coding (RVLC) to a region having a lower priority such as a discrete cosine transform data region (DDP: DCT Data Part) and does not allow the channel characteristic. If not, the low priority region controls the variable length coding means 32 and the inversely variable coding means 34 for variable length coding (VLC). Alternatively, a means (not shown) for inserting additional information is further provided, such that when the channel characteristic identification means 36 determines that the channel characteristic is allowed, such as a discrete cosine transform data area (DDP: DCT Data Part). It is also possible to insert additional information in an area of low ranking. In addition, in the case where the channel characteristic identification means 36 determines that the channel characteristic is allowed, the inversely variable coding means 34 performs inversely variable coding using different coding tables to achieve the effect described with reference to FIG. It is desirable to.

In addition, the video coding apparatus according to the present invention uses a first macro block address (MBA) indicating the number of the first macroblock in each video data packet to achieve the effect as described with reference to FIG. The apparatus further includes means for inserting (not shown), and the inversely variable coding unit 34 preferably inverts the final motion vector value LMVV. In addition, the video coding apparatus includes means (not shown) for inserting a packet number (PN) indicating packet serial number after the resynchronization marker (RM) as packet header information, and a motion vector data area ( MVDP) preferably further includes means (not shown) for inserting the final motion vector value LMVV containing absolute motion vector information.

On the other hand, the marker insertion means 38 is a header data as a resync marker (RM) for distinguishing each video data packet, and information bits for distinguishing the header data area (HDP) and the motion vector data area (MVDP). Header Marker (HM) made using codewords not used for coding of region, and Motion Marker (MM) to distinguish Motion Vector Data Region (MVDP) and Discrete Cosine Transformation Data Region (DDP). Insert). At this time, it is preferable that the header marker HM be easily identified at the time of decoding by using a codeword not used for coding of the header data area HDP.

4 shows an example of a video data packet generated by the method. Referring to FIG. 4, the video data packet generated by the method includes a resync marker (RM), a packet number (PN), and a first macro block address (MBA). do. The video data packet also includes a header data area (HDP), a header marker (HM), and a motion vector data area (MVDP). The video data packet also includes a last motion vector value (LMVV), a motion marker (MM), and a discrete cosine transform data area (DDP: DCT Data Part). In the video data packet thus made, the resynchronization marker RM marks the start of the video data packet. Thus, at the time of decoding, the starting point of the video data packet can be found in the same manner as described in the prior art. The packet number PN indicates the serial number of each video data packet as packet header information. The first macroblock address (MBA) indicates the number of the first macroblock in each video data packet. In the present embodiment, the header data area (HDP) combines macroblock identification (COD) and macroblock pattern chromaticity (MCBPC) to form a new syntax (COD + MCBPC), where COD + MCBPC is a titer. Disguise coding (RVLC). The header marker (HM) is a part that distinguishes the header data region from the motion vector data region and uses a codeword that is not used for COD + MCBPC coding of the header data region HDP. The motion vector data area (MVDP) is a motion vector data portion obtained by inversely distorting coding information generated by motion vector prediction. The discrete cosine transform data area (DDP) contains information related to the discrete cosine transform (DCT), and CBPY, DQUANT, and DCT coefficients are coded. In addition, it is preferable to insert information necessary for error checking and recovery when receiving and decoding from the information used in the corresponding video data packet in the discrete cosine transform data area DDP. Insertion of this information can also be selectively applied depending on the characteristics of the channel. The motion marker (MM) is used to distinguish the motion vector data area (MVDP) and the discrete cosine transform data area (DDP).

A process of decoding the encoded video data packet will be described below. First, when resynchronization marker RM is found in the received bitstream, it is regarded as one packet until the next resynchronization marker RM is found as described in the related art. If an error occurs in the header data area HDP, since the header data area HDP is inversely decoded coded, not only the forward decoding but also the reverse decoding is possible. For example, the header data area HDP may be restored by using the packet number PN and the initial macroblock address MBA of the next packet. That is, since the value obtained by subtracting 1 from the first macroblock address (MBA) of the next packet is a macroblock address of the current packet, it is possible to recover up to a portion where an error occurs in the header data area (HDP) by using reverse decoding. Further, the motion vector data area MVDP is decoded until the motion marker MM is found. Thereby, the motion vector by motion vector prediction is decoded. If an error occurs during the decoding process, since the motion vector data region (MVDP) is also inversely decoded coded, backward decoding is possible. As compared with the process of decoding the video data packet encoded by the conventional coding method, conventionally, since only motion vector data by motion prediction exists in the motion vector data area (MVDP), the motion vector data is previously decoded. It can be used only when there is motion vector data. However, according to the coding method according to the present invention, since the absolute motion vector is unexpectedly transformed and transmitted after the motion vector data region, the previous motion is decoded backward using the final motion vector value (LMVV). Decoding can be performed independently for the vector. In addition, the discrete cosine transform data area (DDP) has no choice but to ignore the video data packet when the variable length is coded when an error occurs. You don't have to. That is, the effect of higher error tolerance can be obtained. In addition, when additional information necessary for reconstruction is encoded and added to the discrete cosine transform data region, it may be used for decoding by using this information.

As described above, since the coding method according to the present invention can further increase the error tolerance, communication is less sensitive to errors in an error-prone environment such as a wireless communication channel.

In the above, the coding method according to the present invention uses the MPEG-4 visual and H.263 as an example, resynchronization marker (RM), packet number (PN), initial macroblock index (FMBI), and header data area (HDP). Terms, such as header marker (HM), motion vector data region (MVDP), motion marker (MM), and discrete cosine transform data region (DDP), have been defined and easily understood by those skilled in the art. It can be applied to any video coding that consists of other equivalent areas that may be.

As described above, according to the present invention, error tolerance can be further increased, so that communication is less sensitive to errors in an error-prone environment such as a wireless communication channel.

Claims (3)

In the video coding method for generating a data packet by coding video data, Coding said video data, The coding of the video data is performed by one of inverse variable coding and variable length coding, and the video coding method comprises information indicating a type of the performed video data coding. A computer readable recording medium executing a video coding method of coding video data to generate a video data packet. Program code for rearranging and data partitioning macroblock data constituting the video data; And And program code for variable length coding or reverse variable coding the divided regions. The method of claim 2, The program code for dividing the data includes program code for generating a discrete cosine transform data region by grouping the discrete cosine transform (DCT) data included in the macroblocks.