Classification Based Inter-Frame Prediction in Video Compression

Video compression is the technique of reducing the original size or storage of the video content. Traditional video coding uses the block-based translation motion model for this purpose. This approach often fails for preserving the good quality of compression, if the motion discontinuities exist in the video frames. A Block that contains the boundary of moving object need more bit for better prediction of the current frame. In the presence of complex motion model the traditional method catastrophically fails to predict the objects having the multiple motion. In this paper, we present a technique for the classification of blocks either it contains the single motion or multiple motion. Experimental result shows that the average accuracy for classification of HD sequences is 85.76% and the achieved bit rebate is up to 1.083% over standalone HEVC.