A new complexity control algorithm for transform and quantization in AVS-M Video encoder

AVS-M is a new video standard used for mobile application. The key of applying AVS-M standard is implementation of efficient coding algorithms. An efficient method is to reduce number of computing transform and quantization by determining all zero blocks before do them. However, complexity reductions by existed algorithms vary with different video content and is disadvantageous for fixed frame rate coding. This results in decrease of video continuity and subjective quality. In this paper, a new algorithm for controlling computation complexity is presented. It controls number of blocks need to be transformed and quantized near to a constant by adjusting determination threshold of all zero block detection adaptively. Simulation results show range of complexity change is no more than 3% and algorithm convergence within 24 frames under acceptable video quality.