Fast and effective CU size decision based on spatial and temporal homogeneity detection

摘要

High-Efficiency Video Coding (HEVC) is the latest video coding standard of the Joint Collaborative Team on Video Coding (JCT-VC). HEVC noticeably improves compression performance when compared with previous standards such as H264, and represents a major step forward in video compression technology. However, this improvement is achieved by increasing the complexity of the encoding process. HEVC employs a novel flexible quad-tree coding block partitioning structure that enables the use of large and multi-sized coding, prediction, and transform blocks. This system is more efficient but also more computationally demanding. In this article an optimized CU size decision algorithm is proposed to reduce the computational cost of quad-tree partitioning by means of spatial and temporal homogeneity analysis and classification, which are directly applied to the input image. If a CU is classified as spatially or temporally homogeneous the quad-tree recursive process is stopped. Furthermore, this image pre-analysis is performed using logic units and embedded hardware on a GPU, thus avoiding unnecessary waiting states, so the computational cost associated with this process is zero for the processor in charge of the encoding process. In comparison with the reference HM16.2 test model, the encoding time is reduced by up to 32.69%, with negligible quality loss and a maximum BD-Rate increase of 1.2% for low-delay P configuration.