Scene adaptive MPEG encoding algorithm using the P-picture-based analysis

Abstract

Presents a novel scene-adaptive MPEG encoding algorithm using parameters encoded in P-pictures. In the MPEG format, a P-picture can have two types of macroblock: intra-coded macroblocks and forward-predicted macroblocks. The proposed algorithm defines a correlation function based on the number of different types of macroblocks and applies this function to detect scene changes in the video sequence. The frame type is then adaptively arranged according to the detection result. Because of motion compensation, the scene change detection is very reliable. Since all the encoded parameters are available during the encoding process, only minimal additional computation is needed. Therefore, this algorithm is faster and more accurate than those approaches using histogram matching.