Research On Fast Inter Mode Decision In HEVC

The new video coding standard, High Efficiency Video Coding(HEVC), was developed to replace H.264/AVC. Compared with H.264/AVC, HEVC adopts more flexiable data structure and achieves high efficiency by introducing a new coding structure in adoption of Coding Unit(CU), Prediction Unit(PU) and Transform Unit(TU). However, it also imposes great computation burden on the mode decision of encoders. So how to reduce the compelixty of HEVC encoder based on ensuring the efficiency becomes a very important problem in academia.In this paper, we performed researches on fast mode decision algorithm and proposed two novel algorithms. The first one is a fast CU depth decision scheme to reduce the encoder complexity for HEVC. This method explored the relationship between rate-distortion(R-D) cost and CU depth and modeled the CU splitting as a binary classification problem with applying Support Vector Machine(SVM) algorithm. The experimental results show that proposed algorithm achieves up to 59% running-time reduction with negligible loss in terms of PSNR and bit rate. The second alogrithm we proposed is CU and PU mode mapping method by precoding with H.264. The input frame is precoded with an efficient H.264/AVC encoder to gather the partition and prediction information of macroblocks. Then two schemes are proposed to map the partition and prediction mode of MB to CU and PU candidate modes. The experimental results compared with x265 show that proposed alogrithm can effectively reduce ~13% encoding time with negligible loss in terms of PSNR and bitrate.