Arithmetic Coding In H.264/AVC And Its Improved Methods

The increasingly wider application of digital video has promoted the continuing improvement of image compression standards. The new standard H.264/AVC has emerged and boasts a series of unique features.It supports the video stream with high quality and low bit rate transmitted over cable, terrestrial and satellite broadcasting. What's more, H.264 could provide services of VOD and stream media on various networks for its flexible network interface. As one of the main features in H.264/AVC, CABAC fully uses the context information in video stream and can be adaptive to the stream, besides, the arithemetic coding has its advantage in computing accuracy, so CABAC can improve coding effiency and has been adopted as one of the two ideas in entropy coding.Though CABAC can improve compression rate a lot, it is very slow in coding speed and needs memories to store the 399 context models.Some special scenes such as DSP based real-time survalliance system have high requirements for compression rate, coding speed and the memory, so CABAC is not qualified and CAVLC, because it's based on VLC whose redundancy can't be reduced completely because of its charteristics, is also not a good option.Because arithmetic coding is good for compression, so this thesis proposes an equalized arithmetic coding framework to substitute CABAC and CAVLC in these scenes.In anther hand, because arithmetic coding is week in error rebustness, it will be a catastropy to the following decoding once there is error in the compressed file. In practical application of wireless channel, because of the unstability in network transmission, it's really possible that the transmission errors will occour a lot, so it's natural that error dedection should be adopted. The existing error detection idea has been proved to be an effective way of fast error detection, so in order to enchance the decoding quality, retransmission or error concealment should be adopted.In this thesis the author concentrates on the following two aspects:1) This thesis improves a fast arithmetic coding algorithm in both coding speed and compression rate. Then designes an equalized arithmetic framework based on the possibility distribution of H.264 syntax elements after binarization to be the substitution of CABAC and CAVLC for scenes having requirements of compression rate, coding speed and memory.This thesis reveals the coding speed bottleneck of CABAC by analyzing its complexity at first, then a fast atithetic coding algorithm is analysied and the improvements both in coding and compression rate is expected.This thesis improves the coding speed by adopting the concept of MPS and LPS and does the periodical statistics of coding symbols to emhance the compression rate. Based on the above improvements and the possibility distribution of H.264 syntax elements after binarization, an equalized arithmetic framework is proposed to separate the coding into two parts adaptive and fixed.The experiments show that it can be used as a substitution of CABAC and CAVLC for scenes having requirements of compression rate, coding speed and memory.2) This thesis analyses the characteristics of CABAC decoder in detail and combines the exsting error detection to propose a retransmission model for application layer, both encoder side and decoder side are processed, so the retransmission data can be reduced and the previous correctedly decoded data can also be reused. The Experiment is used to verify the retransmission model.As a substitution to retransmission, the decoder side can also use error concealment to improve the image quality.This thesis analyses the characteristics of CABAC decoder in detail and combines the exsting error detection to propose a time domain error concealment idea with CABAC error detection. In the experiment,the files with error are processed using error concealments after CABAC error detection and H.264 original code respectively, the performances are compared.Results show that the error concealment idea is effective.