Study On The H.264 Software Codec Algorithm Optimization And Simulation

With the rapid development of network, storage and telecommunication technology, H.264/AVC is applied in broad fields because of its efficient high compression capability and net-friendly performance. However, the rich sets of encoding tools also lead to the increase of complexity. H.264 codec can be implemented on various platforms, and algorithm optimization and simulation are generally considered as one key step in the development. The study of H.264 algorithm optimization, can keep codec with simple procedure, and keep the system in low complexity, and satisfy the application requirement.Based on the H.264 codec scheme, the high complex modules, such as MB mode decision and prediction, integer DCT and quantization, loop-filter, are focused on. In encoding terminal, by modification of a mode decision method based on probability and adding an adaptive controler to set the partitions (subpartitions) of MB dynamically, the mode decision complexity is decreased. The probability feature space is created by three parameters: fintra, which is SATD of MB partitioned by 16 block4x4 with a subset of intra4x4 prediction; finter, the SATD of MB predicted from the previous frame; fmvlength, the MV length of inter prediction. The space is partitioned by train algorithm, with a risk coefficient as input, the coarse intra or inter mode is found if the risk is tolerable. Meanwhile, an AZB detection and skip algorithm is integrated into the scheme. Thus the complexity is lowered with allowable video quality. In decoding terminal the deblocking filter is analyzed and tested because of its complexity.The algorithm scheme based on probability in encoding terminal, decreases the prediction complexity effectively with allowable video quality according to the risk coefficient. Other adaptive controls can be added to this scheme easily. In decoding terminal, some optimization considerations of deblocking filter are verified with tests.