Full rate distortion optimizaton of MPEG-2 video coding

Rate distortion optimization can significantly improve encoder performance in most video coding applications. Taking MPEG-2 as an example, this paper addresses the full rate distortion optimization for the first time by searching the product space of all four free parameters regardless of the computational complexity. An efficient graph-based searching algorithm was extended to the MPEG-2 video coding case from the authors' previous work to find the optimal coefficient indices in the optimization process. The proposed full rate distortion optimization algorithm serves the important role of providing an upper-bound benchmark without considering frame dependence and quantization weight matrix optimization. Two approximations of the full rate distortion optimization algorithms are also presented in order to reduce the computational complexity with a gradual degradation of the rate distortion performance. On average, 2~3dB gain is achieved with the proposed algorithms compared to regular MPEG-2 encoder.