An improved rate control algorithm based on a novel shot detection scheme for the H.264/AVC standard

Abstract

H.264/AVC standard uses a variety of advanced signal processing/video coding techniques for broadcast quality video on half the bit rate required in MPEG-2. Despite being advanced, the current rate control algorithm of the standard does not exploit the fact that typical video sequences are composed of different shots. This paper argues that shot information can help us choose the size of the group of pictures (GOPs) more accurately, thus improving the performance of the rate control algorithm. We initially propose a novel MPEG-7 descriptor driven shot detection method with low computational cost for H.264/AVC. Then we exploit it to determine the initial quantization parameters for each shot based on modelling. These two steps can help us reduce the bit rate and PSNR fluctuation when video sequences have multiple shots. Our proposed scheme outperforms the rate control of the H.264/AVC significantly in terms of reducing the average bit rate fluctuation (variance) by 12.78–99.99% and the average PSNR fluctuation (variance) by 1.4–71% between shots. Experimental results also demonstrate that we can achieve similar or even better rate distortion (R-D) performance than the standard rate control algorithm. Our scheme is also applicable to computationally and memory restricted devices since it needs maximum two frames buffer space for MPEG-7 descriptor calculation, while the average amount of extra processing is only about 5.5% of the total CPU cycles.