Research On Rate Control Techniques For HEVC

As video could give people the most intuitive,rich information,related application obtained great development in the moment.However,as the raw video captured by camera is relatively large,it is necessary to compress before stored and transmitted.The video coding standard provides a unified streaming format and the corresponding technical specifications for the video compression technology,which is the basis for the video industry.A new generation of video coding standard which is called High Efficiency Video Coding(HEVC)was released in January 2013.Although the network bandwidth has gained substantial improvement,it is still not able to provide adequate resources in the process of transmission.Rate control plays a crucial role in the process of video coding,which balances the bandwidth resources and video quality.It helps to adjust the coding parameters so that the quality of video can adapt to the current bandwidth,thus utilizing the bandwidth more efficiently.This dissertation focuses on screen content coding and real-time transmission for rate control and applies them to HEVC.The main innovations and contributions of this dissertation are as follows:Screen content sequences have greatly different characteristics compared to camera captured sequences,which makes exiting rate control schemes unsuitable.While camera captured sequences have relatively good continuity,in screen content sequences,consecutive pictures may present significant movements or abrupt changes.The abrupt changes may consume large amount of bits,which will make bits out of control and affect the subsequence frames.In this dissertation,the characteristics of each frame are modeled and analyzed so that the abrupt changes could be detected effectively.Meanwhile,the compensation window is added after the abrupt picture to guarantee the quality of the picture.Furthermore,a sliding window is introduced to conduct bit allocation instead of GOP level bit allocation.The proposed algorithm has been published in related conference and national invention patents.As users have low tolerance for delay in real-time transmission,the size of virtual buffer will be greatly reduced,which would lead to buffer overflow in the encoding process.This dissertation models the picture classification filter through deep analysis and proposed the virtual buffer management scheme based on several reasonable assumptions.It could predict virtual buffer overflow and improve coding efficiency in real-time transmission.The proposed algorithm has been published in related conference.