Improved Frame Level MAD Prediction and Bit Allocation Scheme for H.264/AVC Rate Control

Abstract

In this paper, we present an improved frame level mean absolute difference (MAD) prediction model for H.264/AVC rate control. Based on the histogram information of successive frames, the MAD of current frame can be accurately estimated. Instead of only using buffer status in the frame target bits allocation process, we also take into consideration the frame complexity which depends on the improved predicted frame MAD value. The objective of this frame complexity based bit allocation is to follow the non-stationary characteristics of video source and to provide smoother visual quality. Simulation results show that by using our proposed scheme, smaller frame target bits mismatch can be achieved. In addition, the average peak signal-to-noise ratio (PSNR) for the reconstructed video is improved by up to 0.5 dB with up to 62% reduction in picture quality variation.