Fast inter-frames pattern chosen algorithm based on H.264
Qiwei Lin, Gui Feng · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2007
As we known that under the same image reconstruction quality, the encoding efficiency of H.264 video standard is about 50% higher than that of H.263 standard. But at the same time, this methodology improving the encoding efficiency based on the rate distortion optimizing (RDO) and the multi patterns motion estimating; as a result it is greatly increasing the calculation complexity of encoder, and greatly affects the video compression speed. Therefore, when we estimate motion pattern of a video sequence, if the best predicting pattern can be decided as quickly as possible, and unnecessary predicting pattern searching can be limited or costly computation can be cut down, then it is possible to speed up the motion estimation. Under these thought, a fast inter-frames pattern selecting and motion pattern estimating algorithm was proposed in the paper. Based on the texture feature and local motion character of current macro block (MB), a set of effective predicting patterns were selected. This selection can effectively reduce searching time of follow up predicting pattern; so it can obviously decrease the operation complexity of the encoder, and improve the searching speed for the motion estimating.