Research On Surveillance Video Coding Optimization For HEVC

The video surveillance system provides great protection for the security of public environment and people's lives.However,the massive monitoring video data puts high demands on storage,transmission and analysis,which poses a great challenge to surveillance video coding.In recent years,surveillance video coding algorithms based on background models have obtained significant improvement.However,the high-quality reconstructed background image will result bursting bit-rate peaks in the video transmission,therefore may lead to buffer overflow and transmission delay.At the same time,there are lot of perceptualredundance that can be compressed in HEVC.According to the characteristics of surveillance video,combined the existing surveillance video algorithms,this thesis optimize the coding in HEVC framework from two aspects of background updating and bit rate allocation.In order to solve the above two issues,the main research content is as follows:Firstly,a monitor video coding method based on background difference is implemented in the HEVC framework.The foreground is separated by the high quality background image,and the foreground part is separately encoded,and the performance gain of the surveillance video is achieved in the HEVC framework.Aiming at the problem of sudden increase of code rate caused by background frame,this thesis proposes a background frame coding algorithm based on luminance compensation algorithm,The proposed algorithm firstly adjust the current background frame's color by the luminance compensation algorithm,then predict the background frame by a novel adaptive skip prediction scheme,finally encode and transmit the rest residuals.Experiments on the PKU-SVD-A database showed that the proposed approach not only improves the coding quality,but also smooth the bit-rate peaks efficiently,which are caused by the transmission of background frames in surveillance video coding.Secondly,the code rate control algorithm used by the existing HEVC test model HM16.7 is studied and analyzed,which is combined with the background frame coding algorithm proposed in the previous text.Aiming at the unreasonable bit allocation in the CTU level of bit rate control,an improved CTU bit allocation algorithm based on foreground performance gain is proposed.The experimental results show that the proposed method achieves effective rate control,and maintains the coding advantage based on the background model,at the same time,under the condition of maintaining the constant bit rate,the low delay P frame coding mode,the Y-PSNR average of the foreground area are improved by 0.36 dB,and the background area and the overall Y-PSNR have a slight decrease.