Research On Temporal Resolution Enhancement In H.264 Video Compression

Video reconnaissance is one of the most important reconnaissance means in modern war, but the contradiction between the large data of the video signal and the low transmission rate of the recon system restricts the development of video surveillance systems, and therefore high real-time video compression system becomes a research focus in the video recon field. As the compression will result in the loss of information of the original video sequence, so how to enhance the information missed during video compression in the post processing becomes another issue to be resolved.H.264 video compression standard has a higher coding efficiency than those existing video compression standards. It can reduce half of the bit-rate with the same quality. Coupled with low latency and adaptability, H.264 is very suitable for the real-time video compression system. Based on the summary of the development of the digital video compression technology, the frameworks and main techniques of H.264, such as prediction coding, transform coding and entropy coding, are analyzed in the thesis. And then a video compression system which is based on the TMS320DM642, is realized. This system uses DSP+FPGA hardware architecture, and makes full use of the computing and data control capabilities of the both. As the transmission bandwidth is limited, a temporal domain down-sampling approach is introduced to the system to reduce the transmission bit-rate. Therefore, the receiver needs to reconstruct the missed frame to improve the temporal resolution of the video sequence. We research on the algorithm of the temporal domain interpolation that is base on the motion estimation in this thesis. Motion estimation algorithm is studied in the first, and proposing an adaptive quasi-cross search method for motion estimation based on the classic algorithm. Then, we study several temporal resolution enhancement algorithms, propose improvement for the algorithms. Finally, carrying out experiments and analyzing the result.After commissioning, the video compression system works stably, which can compress and decompress CIF image at the rate of 20 frames per second simultaneously. The system can achieve good restoration images at the compression rate of more than 20. Improved Temporal Interpolation algorithm works well, and can achieve good reconstruction images in different levels of temporal resolution enhancement.