Histogram-based rate-distortion estimation for MPEG-2 video

摘要

A rate-distortion estimation method for MPEG-2 video that enables one to predict the number of bits and the distortion to be generated from an encoded picture at a given quantisation step size, and vice versa is proposed. The estimation method has some outstanding advantages: first, the computational complexity is small because its major operation is just to obtain a histogram or weighted histogram of the DCT coefficients from an input picture, and the final form of the proposed estimation model is simple. Secondly, its results are accurate enough to be applied to practical video coding applications. Simulation results show that the estimation errors for the rate and for the distortion do not exceed 2.5% and 1%, respectively.