Research On H.264 Motion Estimation Algorithm And Application Of H.264 Encoder On Davinci Platform

Compared with the previous video coding standard,H.264 introduces some new techniques,such as multi-reference frame,variable block size,fractional pixel motion prediction and rate control.Under the condition of the same quality of the reconstructed image,a higher compression ratio can be achieved by H.264.H.264 becomes the mainstream video coding standard for its higher compression ratio,higher image quality,lower rate,good network affinity and other characteristics.In the H.264 encoder,the computational complexity of the motion estimation using the full search method is more than 80% of the whole coding complexity.Therefore,the improvement of the section of the motion estimation is the research focus about the H.264 encoder.In addition,with the increasing demand of the video conference,security and other aspects,the application prospect of the H.264 encoder in embedded platform is very broad.In recent years,many H.264 fast motion estimation algorithms have been proposed.These algorithms can be classified into three categories,which are based on the idea of template matching,the dynamic search range,and the reasonable search strategy separately.These fast motion estimation algorithms can significantly reduce the time consuming of the motion estimation in the case of the same quality of the reconstructed image.Accordingly,the real-time performance of the H.264 encoder is improved in the embedded platform.In this paper,the key techniques of the H.264 encoder and some classical motion estimation algorithms are analyzed.In addition,the focus of the research content also includes the fast motion estimation algorithm recommended by the H.264 standard,Unsymmetrical Cross Multi Hexagon Grid Search(UMHexagonS).The search speed of the UMHexagonS algorithm is very slow.To solve this problem,an improved fast motion estimation algorithm is proposed based on the relative properties of the predicted motion vector.There are three improved aspects,the prediction of the initial search point,the global search and the early termination strategy.Experimental results show that the proposed algorithm can adapt to the video sequences with different intensity under the condition of different quantization steps.This proposed fast motion estimation algorithm is suitable for H.264 with good performance.Based on the above,the H.264 encoder is applied and optimized on the DaVinci platform.Experimental results verified the efficacy of the transplantation and optimization of the encoder.