Error Resilience Technique For Variable-length Code Block

The development of network technology takes this tendency a step further. In order todeal with mass data transmission in the network, experts and scholars make great efforts toimprove the efficiency of data transmission and storage by setting a series of videocompression coding standards successively, such as the MPEG-x series and H.26X seriesstandards, which have been widely used in many fields. However, the improvement ofcompression efficiency raises the sensitivity of code streams to the error rate and calls forhigher demand of data robustness. Hence the error resilient algorithm is improved in thispaper. This algorithm is applied to H.263video coding standard with full comprehensiveconsideration of the encoding complexity and running time. The related contents will beintroduced as the following aspects:(1) The error resilient entropy coding is improved. The traditional algorithm turns thecode stream into equal-length code stream by filling the different length of data blocks intounequal-length data slots according to certain rules. Although this approach has improved therobustness to a certain extent, it is susceptible to noise interference due to the weak protectionfor low frequency coefficients. The new algorithm groups the equal-importance coefficientstogether and sorts the data bits by there importance. Compared with the block data structure intraditional method, the low frequency coefficients is protected by the data structure in ourproposed algorithm with rate control through unequal error protection. Due to the eliminationof the correlation between bits in one data block, the diffusion of bit error is retarded.(2) The implementation of H.263video codec. The H.263video compression standardis established by ITU-T to apply low bit rate video transmission. Based on some efficientfunction of Intel IPP library, this thesis implements the H.263codec, which includes intracoding and inter coding. The intra coding consists of Discrete Cosine Transform (DCT),quantization and variable length coding. The inter coding consists of estimation andcompensation.(3) The codec is tested on VS2008platform. Firstly, the reconstructed images underdifferent quantization levels are compared. Then the effect of different bits on the decoding is tested under different parameters. At last the reconstruction results of three video sequencesunder different bit error rates are tested to show the difference of the original algorithm andthe improved algorithm. The video frames are reconstructed and the relevant comparisoncharts are drawn in MATLAB. The experimental results show our proposed algorithmimproves the peak signal to noise ratio (PSNR) by1~2dB than the original algorithm.