Fast Motion Estimation And Intra Mode Prediction For H.264

As a new generation of video coding standard, there are many technical innovations in H.264. Whether it is used for storage or network transmission, H.264 is competent with a higher compression ratio, better quality of compression image and network compatibility. However, the algorithm complexity of H.264 dramatically increases due to the adoption of a number of video coding tools. So how to reduce the complexity so that it can meet the requirements of real-time communication has become an urgent problem that needs to be resolved.H.264 is still an algorithm based on block-matching motion estimation to remove temporal redundancy. A fast search method-adaptive threshold search algorithm is proposed in this paper based on the analysis and research of several traditional fast search algorithms. It takes advantage of spatial-temporal correlation of the motion vector field to predict the initial point and obtain the adaptive threshold, and terminates current search immediately once detects the stationary block. And according to the characteristics of the motion of image, different templates are adaptively utilized. The algorithm is implemented in the test model JM8.6 and compared with the fast algorithm of test model. The experimental results show that the algorithm in this paper improves the motion estimation's speed to a certain extent while ensures the search accuracy.Intra prediction is a key technique in H.264.It predicts the block with a variety of models,and selects the model with best rate-distortion performance as the the best predictive model of current block. It also increases the computation burden when improves the performance. After introducing the mode selection algorithm of intra prediction of H.264 and analysing several fast algorithms for intra mode selection, an improved algorithm is proposed by using the image texture features in this paper.It narrows the select scope of intra-prediction according to different texture features ahead of time. The experimental results show that it improves the encoding speed to a certain extent while ensures the image quality.