JP5474896B2 - Video error detection technology using CRC parity code - Google Patents

Description

  The present invention relates to an error detection technique in a video bit stream.

  MPEG, H.M. Implementations of current video compression standards such as H.263, JVT (also known as ISO / ITU H.264) divide video frames into macroblocks and slices. A slice generally includes a given number of consecutive macroblocks. However, using the Flexible Macroblock Ordering option for videos that are compressed according to the JVT compression standard can result in macroblocks of unwanted slices that are adjacent to each other in display order. Variable length coding (VLC) technology can improve compression efficiency at the expense of error tolerance. Bitstream errors can propagate undetected and adversely affect all subsequent macroblocks in the slice.

  To allow error recovery, the header of each slice includes a resynchronization marker. Such a marker constitutes the earliest point at which proper decoding can occur following a bit error. Slice headers start on byte boundaries so that they are byte aligned, and each macroblock is not byte aligned. Using context-based adaptive binary arithmetic coding for encoding video bitstreams according to the JVT compression standard, multiple symbols can be encoded as one bit, so the macroblock boundaries are not even bit aligned . This makes it difficult for the decoder to accurately specify when a bit error occurs for a variable-length codeword. Most bit patterns in the affected bitstream represent valid codewords even though they do not represent the exact value. In some cases, the identification of codewords that are not valid can occur when the decoded value exceeds the normal range, but this is always variable and up to an unknown amount of incorrect encoding. It may not be obvious. Thus, all macroblocks in the slice following the error remain potentially affected.

  Video error detection can be performed by decoding the video frame and then checking the frame for spatial discontinuities at the macroblock and slice boundaries in the pixel domain. There are various proposals for video error detection and concealment. Following error detection in a particular macroblock, error concealment can be performed on the affected macroblock. A concealed macroblock generally looks perceptually better than an affected macroblock that is not concealed, but is perceptually worse than a properly decoded macroblock. However, current concealment algorithms are computationally intensive and use significant resources. In addition, such algorithms also tend to be false positives (conceal properly decoded macroblocks) and false negatives (do not conceal affected macroblocks).

  One approach to solve the affected macroblock problem is to transmit the entire slice in one IP packet. If a slice disappears during transmission, all macroblocks in the slice are marked as missing, which requires concealment. As described above, all data in a packet due to a known transmission error is marked as affected, discarded and concealed. For this reason, the entire slice will disappear in the case of the detected transmission error. Assigning a relatively small number of macroblocks per slice improves error resilience, but at the cost of significant coding and transport efficiency.

  Therefore, a technique for detecting a bitstream error with a relatively small computational complexity is required.

  In summary, according to a preferred embodiment, a method for detecting errors in an encoded video stream is provided. The method starts by receiving an encoded video stream with a first parity code calculated for a group of N macroblocks of the stream before transmission. The stream is decoded to form a group of decoded macroblocks. A second parity calculation is performed on the decoded group of macroblocks, and a comparison is made between the first and second parity codes to determine whether the decoded macroblock group has an error. .

  In summary, according to another preferred embodiment of the present principles, a technique is provided for detecting errors during the transmission of a video bitstream composed of compressed macroblocks. A CRC (Cyclic Redundancy Check) parity code is calculated for a group of compressed macroblocks that form a video stream prior to transmission. Upon receiving a group of compressed macroblocks, a second CRC parity calculation is performed. These CRC codes are compared. If the CRC codes match, no error has occurred and normal decoding of the group of blocks and other processing can be performed. When an error occurs, the group macroblocks are error concealed. When a group of macroblocks is compressed according to the JVT video compression standard, a CRC parity code can be transmitted as an SEI message.

  According to the present invention, it is possible to provide a technique for detecting a bitstream error with a relatively small computational complexity.

FIG. 1 shows a block schematic diagram of an exemplary embodiment of a transmitter according to a preferred embodiment of the present invention that performs a group of compressed macroblocks and an accompanying CRC parity calculation. FIG. 2 shows a block schematic diagram of an exemplary embodiment of a receiver that receives compressed macroblocks from the transmitter of FIG. 1 and performs CRC parity calculations to perform error detection and concealment as needed. .

1 and 2 each show a transmitter according to the invention that works together to detect errors in a video stream transmitted from a transmitter to a receiver and to conceal such errors when it is detected. 10 and the receiver 20 are shown. As will be described in detail below, the combination of the transmitter 10 and the receiver 20 uses CRC (Cyclic Redundancy Check) to detect an error. The CRC constitutes a commonly used error detection code used in generating an FCS (Frame Check Sequence) attached to a block of data before transmission. Upon receipt of the data block, the CRC enables bit error detection. CRC codes are usually applied to 8-bit, 16-bit and 32-bit symbols. One commonly used 16-bit CRC generator polynomial is X ¹⁶ + X ¹² + X ⁵ +1.

  The number of bits or bytes to which the CRC is applied must be known when received in order to perform error detection. CRC is applicable to fixed-length data blocks, and a predetermined fixed length is known at the transmitter and receiver. Alternatively, the CRC can apply a variable length packet, which has a length field in the header so that the receiver can determine the length to which the CRC check calculation should be applied. When the CRC check at the time of data reception indicates an affected packet, the CRC does not provide information about the location within the packet where degradation has occurred, so even if there is a valid data portion, generally all the data in the packet Is destroyed.

  To solve the above problems, the transmitter and receiver of FIGS. 1 and 2 each associate an association between each macroblock and the compressed data bits or bytes associated with the macroblock. When compressed data representing a macroblock is not aligned by bits or bytes, a partial bit or byte bit or whole byte is associated with the macroblock. One bit or one byte may be associated with multiple macroblocks. As described in detail below, a CRC parity calculation is applied to the compressed bits or bytes for a group of N macroblocks that generate a CRC parity code, where N is an integer. If the next N macroblocks in the slice, or fewer than N macroblocks remain in the slice, then a smaller number of macroblocks are grouped, other CRC parity calculations are applied to them, and other parity codes Might generate. This grouping of N macroblocks and calculation of CRC parity on each set of N blocks continues until all macroblocks of the slice are used. By calculating the CRC parity code on the block set in the slice (not for the entire slice), processing (ie, decoding) is performed for each group of N blocks detected as error-free. Concealment is performed for each group detected as having an error. This avoids the need to discard the entire slice if an error is detected.

  FIG. 1 shows details of the transmitter 10. A video encoder 12 in the transmitter receives the input video stream and includes, but is not limited to, MPEG, H.264, and the like. The video stream is encoded (ie, compressed) according to any of several well-known compression techniques, including H.263 and JVT. In the illustrated embodiment, the encoder 12 encodes the input video stream according to the JVT compression standard. The packetizer 14 packetizes the encoded video from the encoder to the receiver 20 of FIG. 2 for transmission over the network 15.

  A selector block 14 within the transmitter 10 selects N compressed macroblocks in each slice to be encoded by the encoder 12. The selector 14 groups such remaining macroblocks into groups until only less than N macroblocks remain in the slice. The CRC parity calculator 18 calculates a CRC parity code for the bits or bytes that represent the N compressed macroblocks. The calculated parity code is transmitted along with the compressed video data, either by a packet separated from the compressed video data of the slice, or at the beginning of the same packet as the compressed video data of the slice. The value of N is predetermined and known at both the transmitter 10 and the receiver 20 of FIGS. 1 and 2, respectively, or can be transmitted with a CRC parity code. When the video encoder 12 encodes the input video according to the JVT compression standard as in the illustrated embodiment, the CRC parity calculation performed by the CRC parity calculation unit 18 is supplemented with supplemented enhancement information (Supplemental Enhancement Information) along with the encoded video. ) Can be transmitted.

  FIG. 2 shows details of the receiver 20. The receiver 20 includes a video decoder 22 that decodes encoded video packets transmitted over the network 15 by the transmitter 10 of FIG. In the illustrated embodiment, the video decoder 22 within the receiver 20 of FIG. 2 operates to decode the packetized video that was originally encoded according to the JVT video compression standard. The CRC parity calculation device 24 is configured so that each of the N macroblocks decoded by the video decoder 22 corresponding to each of the N decoded macroblocks for which the CRC parity calculation device 18 of FIG. A CRC parity code is calculated for the group. That is, each of the CRC parity code calculation devices 18 and 24 of FIGS. 1 and 2 calculates a CRC parity code for the same group of N macroblocks.

  The CRC parity comparison device 26 in the receiver 20 of FIG. 2 compares the CRC parity code calculated by the CRC parity calculation device 24 with the CRC parity code received from the CRC parity calculation device 18 by the SEI message. The match of the two CRC parity codes indicates that the group of N macroblocks decoded by the decoder 22 contains an error for the same group of N macroblocks transmitted by the transmitter 10 of FIG. Means not. Under such circumstances, the selector 28 allows an error-free group of N macroblocks to be passed to their destination, which may include a display (not shown), a recording device, and the like. If the CRC parity codes do not match, as occurs in the error event, concealment is performed by a video error concealment block 30 that alerts the video decoder 22 of such action for a group of N macroblocks.

  Effectively, the decoder 22 of FIG. 2 knows based on the information received from the video error concealment block 30 which groups of macroblocks are affected by the error. Macroblocks related to each group of slices received prior to the group having an error are normally decoded because such a block has no error. Later macroblocks in the slice that are in error are considered degraded and are concealed instead of decoded.

  Compared with the prior art method in which one CRC code is applied to the whole slice, the error detection technique of the present principle is such that the group of macroblocks preceding the transmission error has no error and the concealment concealment or Provides the effect that it does not have to be a test. Undegraded macroblocks that are improperly concealed generally have lower image quality than properly coded macroblocks. Due to inter-frame coding, low image quality may last longer.

  The error detection technique of the present principle is preferable to the method of applying the spatial continuity index to the decoded video in order to detect the position of the transmission error. It has the effect that the computational complexity is greatly reduced since the calculation of the CRC parity code and the comparison of the received values are subject to much lower complexity than the measurement of spatial discontinuities. The technology of this principle has the effect of avoiding false positives where undegraded macroblocks are improperly concealed, and avoiding false negatives that are not concealed but should be done for degraded macroblocks Has the effect of

  When applied to the MPEG-2 video compression standard, a CRC parity code and optionally a value of N can be transmitted in the user data field. When applied to the JVT video compression standard, a CRC parity code and optionally a value of N can be transmitted by one or more SEI messages or user data.

  The number of macroblocks for which N is selected and the CRC parity code is calculated depends on the expected error rate of the network 15. Selecting a large value of N (the smaller the group of macroblocks, the more macroblocks per group) reduces the CRC parity code transmission overhead, but reduces the accuracy of error detection. Selecting a small value of N (the larger the number of macroblock groups, the smaller the number of macroblocks per group) increases the overhead of transmitting the CRC parity code, but improves the error identification accuracy. For a small value of N, assuming a uniformly distributed error and uniform compression for all macroblocks in a slice, the ratio of macroblocks that can be identified as not degraded approaches ½.

  Although the error detection technique described above uses a CRC code, any type of error detection or correction polynomial can be used. CRC codes are generally calculated and applied to bytes. Furthermore, the present technique does not require CRC parity code calculation and transmission for all macroblocks of a slice. In some examples, sending a parity code for a macroblock that precedes the macroblock after the slice is corrected for the remaining macroblocks of the slice once an error is detected in the compressed slice. Since it is not, it turns out that it is effective. This continues to be true even if they belong to the macroblock group received in transmission error.

  The above describes a technique for detecting errors in an encoded video stream.

10 Transmitter 20 Receiver

Claims (9)

A method for detecting errors in an encoded video stream, comprising:
Receiving the encoded video stream comprising at least one slice having at least a first group and a second group of N macroblocks, wherein the at least one bit of the encoded video stream or bytes are associated with both the first and second groups of the N macroblocks of said at least one slice, the step of receiving,
Receiving a first parity code according to the first group of the N macroblocks of said at least one slice,
Wherein with respect to said first group of N macroblocks of at least one slice, a steps of selecting a bit or byte from the encoded bit stream, the bit or byte entirely the N macro The selecting step wherein the entire bit or byte is selected for the first group of N macroblocks when associated with both a first group of blocks and a second group ;
Calculating a second parity code for the selected bits or bytes;
Comparing the first parity code and the second parity code to determine whether the first group of the N macroblocks includes an error;
Having a method. The method of claim 1, further comprising concealing the first group of N macroblocks when it is determined that the error exists in the first group of N macroblocks.   The method according to claim 1, wherein the first parity code includes a CRC (Cyclic Redundancy Check) code. An apparatus for detecting errors in an encoded video stream,
(A) the encoded video stream consisting of at least one slice having at least a first group and a second group of N macroblocks , wherein at least one bit of the encoded video stream; Receiving the encoded video stream, wherein bytes are associated with both a first group and a second group of the N macroblocks of the at least one slice; and (b) of the at least one slice. receiving a first parity code according to the first group of the N macroblocks, (c) the first group of the N macroblocks of said at least one slice, the encoded bit stream from a receiver to select a bit or byte, the whole bit or byte of N Entire bits or bytes when the first group of macroblocks and are associated with both the second group is selected for the first group of the N macroblocks, and said receiver,
A computing device for calculating a second parity code for the selected bits or bytes;
A comparator for comparing the first parity code and the second parity code to determine whether the first group of the N macroblocks includes an error;
Having a device. 5. The apparatus of claim 4, further comprising a video error concealment device that conceals the first group of N macroblocks when it is determined that the error exists in the first group of N macroblocks. (A) an encoded video stream comprising at least one slice having at least a first group and a second group of N macroblocks , wherein at least one bit or byte of the encoded video stream Receive the encoded video stream associated with both a first group and a second group of the N macroblocks of the at least one slice ; and (b) the encoded video stream a selection device for selecting a bit or byte for the first group of N macroblocks, the whole bit or byte is associated with both the first and second groups of said N macroblocks from The entire bit or byte is selected for the first group of N macroblocks. Said selection device to be selected;
A transmitter for transmitting the encoded video stream together with a parity code calculated for the selected bits or bytes;
Having a device.   The apparatus of claim 6, wherein the parity code comprises a CRC (Cyclic Redundancy Check) code. Receiving an encoded video stream comprising at least one slice having at least a first group and a second group of N macroblocks, wherein the at least one bit of the encoded video stream or bytes are associated with both the first and second groups of the N macroblocks of said at least one slice, the step of receiving,
And selecting a bit or byte for the first group of the N macroblocks from said encoded video stream, the overall bit or byte first and second groups of the N macroblocks The selecting step wherein the entire bit or byte is selected for the first group of the N macroblocks when associated with both
Transmitting the encoded video stream together with a parity code calculated for the selected bits or bytes;
Having a method.   9. The method of claim 8, wherein the parity code comprises a CRC (Cyclic Redundancy Check) code.

Images