Adaptive Simplification of Prediction Modes for H.264 Intra-Picture Coding

Abstract

In an H.264 coding system, an intra prediction method that uses neighbor pixels within the current picture has been adopted for improving coding efficiency. However the large number of intra prediction modes that must be searched significantly increases the number of operations at the encoder. In this paper, we first investigate the usage distribution of intra prediction modes. Then for the focus region of the picture being coded, all of prediction modes are applied. For the other regions, only the important modes are utilized. Important modes are determined from the usage distribution in the previous coded picture; prediction modes that are utilized more than 10% of the time are considered important. From the simulation results, we can verify that the H.264 encoding time may be reduced by approximately 28% for I-picture only and 7% for I, P-pictures without significant loss in the subjective picture quality.