Research And Implementation Of Adaptive Motion Estimation Algorithm Based On H.264/AVC

H.264/AVC is the latest video codec standard nowadays, and it wasproposed by the experts from both ITU and ISO/IEC. Like the videocoding and decoding standard before, H.264/AVC also takes the hybridcoding scheme. But at the same time, it adopts a lot of new techniques.The efficiency of H.264/AVC is greatly improved because of these newcharacteristics. Besides these new features, the codec complexity ofH.264/AVC is also greatly enhanced: about5to10times in coding and2to4times in decoding; Motion Estimation and Compensation is one ofthe most important parts in video coding, and it has a significantinfluence on the performance of the encoder. For example, the timeconsumed in Motion Estimation takes60%-80%percents of the totalencoding time in Full Search, and it cannot meet the requirement of theapplications which have a high real-time demand. That is why developingefficient new fast Motion Estimation algorithms become one of the mostpopular fields among the researchers all over the world.In this paper, a brief introduction to H.264/AVC was given first, andthen some classical and modern motion estimation algorithms werecompared; The theory of motion estimation, the statistical distribution ofmotion vectors and the correlation between the frames of the videosequence were analyzed in the following; At last, an adaptive motionestimation algorithm was proposed based on UMHexagonS. Becausemost video sequences have a smooth or slow moving background, so athreshold was given to terminate the search progress when the differencebetween current block and reference block is less than a specifiedthreshold; In the other hand, the5x5search pattern was also optimizedbased on the statistical distribution of motion vectors; the directional ofmoving objects in the sequence was also taken into consideration toreduce the search points in the multi-Hexagon search. The results ofexperiments show that the algorithm proposed in this paper can improvethe performance of the encoder efficiently with little loss in PSNR.