Efficient Intra Prediction in H.264 Based on Intensity Gradient Approach
An‐Chao Tsai, Anand Paul, Jai-Ching Wang, Jhing-Fa Wang · 2007
This study presents an intensity gradient approach to intra prediction in H.264 encoding system, which enhances the performance and efficiency by means of edge orientation of the gradient filter. We propose a pre-processing stage in which eight-orientation feature are extracted from a macro block that selects four modes to be applied to the block among a set of predefined modes. It is shown that by choosing number of modes used in rate-distortion calculation lead to significant enhancement in the performance of intra prediction. The validity of this algorithm is confirmed experimentally. The simulation results indicate that the intensity gradient-based algorithm for intra prediction contributes to the bit-rate reduction compared to that of previous algorithms and saves around 53% of the total encoding time in H.264.