Fast Intra/Inter Mode Decision Algorithms For H.264

H.264/AVC (Advanced Video Coding) is the latest video coding standard developed by ITU-T and MPEG, which at the same time as the MPEG-4 Part 10. H.264 use a variety of new technologies in the intra-coding, interframe prediction, transform quantization and entropy coding, substantially increased the coding efficiency. H.264/AVC more than doubled the video quality in use with the method similar to the best coder in the same video bit rates compared with H.263v2 (H.263+) and MPEG-4 Simple Profile.In H.264, mainly use rate-distortion optimization algorithm based on the rate-distortion theory to decision different modes. Rate-distortion algorithm which chooses the best mode called the RDO (rate-distortion optimization). While H.264 using RD-OPT algorithm can significantly improve the SNR and bit rate, but it also led to the sharply rising of computing complexity. Judging from the encoding time, the RD-OPT algorithm time is increasing dramatically. In intra-prediction mode, in order to determine a macroblock's intra-prediction mode, 592 kinds of RD_Cost need to calculate. It will bring greater computing complexity in the inter-frame prediction as it relates to the selection of multiple reference frames.This paper first introduced the key technologies used in H.264 encoding standard. Briefly introduced rate-distortion optimization algorithm principle and the mode decison algorithm in H.264 application based on JM. Research and analysis focusing on the intra-/inter-frame mode decison algorithm and describe the utilization of mode decision algorithm, which reduce the computing complexity, improve the encoding speed. This innovation lies in two aspects: First in intra-mode decision, develop a pre-computing algorithm which use of the spatial correlation of macroblock itself to forecast the prediction direction, then reducing the number of decision modes; the second in inter-frame decision, use of edge detection operator to calculate the edge amplitude and the macroblock is classified as two kind, each kind has different modes set, which effectively reduces the computing complexity in inter-frame mode decision.