Weight Based Fast Mode Decision for H.264/AVC Video Coding Standard

Abstract

H.264/AVC video coding standard outperforms former standards in terms of coding efficiency but at the expense of higher computation complexity. Of all the encoding elements in H.264/AVC, inter prediction is computationally most intensive and thus adds to the computational burden for the encoder. In this paper, we propose a fast inter prediction algorithm for JVT video coding standard H.264/AVC. All images have certain characteristics that are inherent to them. Natural videos have many homogeneous regions. In video sequences, there are stationary regions between frames and regions with moderate to complex motion. From an in-depth analysis of the full search algorithm in which properties of each macroblock (MB) are studied in details, a method is proposed where each MB is given a weight depending upon its characteristics. The motion content and the homogeneity parameters of each MB is determined prior to the ME process. The MBs correlation with neighboring MBs in respect of predicted motion vectors (MV) and encoding modes are studied. Weights are assigned for these parameters and the final mode is selected based upon these weights. We propose a Weight Based Fast Mode Decision (WBFMD) process where we define four weights for each MB based on its motion content, homogeneity, value of the predicted MV and the encoding modes of the neighboring MBs. Smaller weights are assigned when these parameters have low values indicating a simple MB whereas larger weights are assigned when these parameters have high value indicating a more complex MB. The average video encoding time reduction in the proposed method is 70% compared to the JVT benchmark JM12.4 while maintaining similar PSNR and bit rate. Experimental results for various test sequences at different resolutions are presented to show the effectiveness of the proposed method.KeywordsH.264/AVCInter predictionMotion estimationWeighted prediction