Fast PU decision algorithm based on texture complexity in HEVC

The HEVC standard adopts varies of Prediction Unit (PU) modes to perform motion estimation (ME). And the mode with the least rate distortion (RD) cost is chosen as the best mode after transversally comparing all candidate modes, which can achieve the highest coding efficiency but require a high computation complexity. In this paper, a fast PU decision algorithm is proposed. Firstly, the correlation among spatial-temporal PU modes is analyzed and its relative adaptive weights are calculated. Next, a classification of CUs is provided based on texture complexity. In each category, the unnecessary PU modes selected is excluded to reduce computational complexity. Experiments show that the algorithm can greatly reduce the coding complexity by averagely up to 45% in the random access configuration with only 2% bit-rate increment and a 0.08db PSNR decrement compared to the one of original HEVC encoder.