| With the continuous improvement and maturity of the new-generation video codec standard HEVC/H.265,its efficient encoding performance has been gradually applied to high-definition and ultra-high-definition video scenes.Meanwhile,the bandwidth capacity of network communication always meets its limitation.Therefore,a unique scheme has been designed to guarantee the visual quality requirements of the pictures under restricted bandwidth,which is called rate control technology.For a long time,rate control has become an important research issue for the academic community.The arrival of big data era has made the collection,storage and application of surveillance video more and more important.The real-time codec transmission of surveillance video for network communications possesses strong practical requirements.Due to the lack of particular optimization strategies for surveillance video coding,various existing standard technologies have caused the problem of video stream transmission applicability.Inter prediction technology is a major method to remove video time domain redundancy,and has always been the focus of video coding research.The existing inter-frame prediction technology is increasingly perfect.How to further develop new methods based on this,and continuously improve the coding efficiency,is facing great challenges.This article focuses on video coding technology and conducts research in three areas.(1)Study dynamic bit rate controls algorithms in instant messaging video.Fuzzy information of image quality is introduced into the video coding,the quality change before and after the image coding is analyzed,the reference image motion information is calculated,and the code rate control model is established to allocate the coding bit rate.A dynamic rate control algorithm based on image ambiguity,objective quality of the reconstructed image and motion intensity of adjacent reference images is proposed,which saves the bit rate while ensuring subjective quality.(2)Research monitoring video scene,coding optimization methods.For controlling the bit rate consumption of the background image,the problem of flushing the code stream during network transmission is solved.A new long-term reference structure is proposed to meet the long-term characteristics of surveillance video coding in inter-frame reference.At the same time,quantization parameters are optimized globally for various frame types.While ensuring the smoothness of the code stream and reliable transmission,the performance of the surveillance video coding performance has been significantly improved.(3)Study the inter-prediction pixel compensation pixel re-correction technique.Using multivariate linear regression method to obtain the compensation relationship between the reference block and the source block,a reference pixel weight correction algorithm is proposed.By improving the quality of the reference pixel block to enhance its reference value,the purpose of improving inter-frame prediction coding efficiency is achieved.Combining the above three aspects of research,the proposed instant messaging video rate control algorithm is directly applied to commercial video encoder x265 to achieve efficient variable bit rate control.The research results on the optimization of surveillance video coding have been integrated into the AVS2 reference software RD,the second generation video standard of Chinese audio and video standards,as part of the standard.The proposed inter-frame reference compensation pixel weight correction technology has been implemented in the HEVC reference model software HM.Tests show that this technology has a strong development potential and will be ported to the next generation video coding standard H.266 software platforms in the future. |
PDF Full Text Request