Improved Fast Mode Decision Algorithms Based On HEVC Screen Content Coding

In this project,fast mode decision algorithm based on HEVC screen content encoding is studied.HEVC(high-efficiency video coding standard)is the new generation of video coding standard.Compared with the previous video coding standard H.264,the HEVC standard can reduce the bit rate of encoded video by 50%while ensuring the same video coding quality.In order to achieve such a coding performance improvement,the HEVC standard introduces many new coding techniques,including more flexible coding unit and more accurate mode prediction algorithms.However,the improvement of compression performance should be based on the computational complexity of encoder.Especially in the process of mode decision,the number of modes that need to be predicted by the encoder increases a lot,which leads to a significant increase of coding complexity.Screen Content Video is a relatively specialized category of video that is typically computer generated and usually contains fewer different shades of color,sharper edges,repetitive graphics modes,irregular play fields and less complex local textures,also sudden scene switch.Screen content encoding technology,is the latest encoding technology to join the HEVC standard in 2016,greatly improving the coding performance of HEVC on the screen content.The fast algorithm proposed in this project aims to optimize the mode decision algorithm for screen content coding.Using the temporal correlation between video sequence frames,this project proposes an improved algorithm,"early termination of CU partitioning and PU prediction algorithm".The algorithm uses the temporal correlation of CU partition depth and PU partition mode,selects the best prediction mode for the current CU,and ends the mode decision process in advance.In order to ensure the rate-distortion performance of the algorithm,we set two thresholds of rate-distortion penalty for the algorithm.The algorithm is implemented in the HM-16.12 + SCM-8.2 test model and compared with the original algorithm.The experimental results show that compared with the original algorithm,the fast algorithm can reduce the coding time by 24.46%and 22.6%respectively when the encoder uses lwdelay mode and randomaccess mode to encode.This project also propose a new efficient algorithm,"U-based fast coding algorithm based on content characteristics".The algorithm uses the distribution of luminance values of pixels in the image,and divides the coding CU into SSC_CU and CSC_CU,and proposes fast coding methods for the SSC_CU and the CSC_CU respectively.The algorithm is coded in the HEVC test model HM-16.12 + SCM-8.2 and compared with the original encoder.Experimental results show that the fast algorithm reduces the coding time by 28.32%when the encoder uses intra-only mode to encode.