Research On Traffic Surveillance Video Coding Technology Based On Virtual Reference Frame

Traffic video denotes image sequence that recorded and saved by traffic monitoring system with monitoring equipment.With the development of urban construction and social civilization progress,continuous expansion of city size, constantly improving of transportation system, continuous improvement of intelligent monitor systems, more cameras installed on the road, lots of traffic videos recorded and saved every moment. Compared with natural images, traffic videos have special features such as significant difference between foreground and background, little changing with light and background substantially retained, which poses a challenge for the coding efficiency of existing image/video coding standards on traffic video compression. Considering the importance of traffic video,seeking efficient compression schemes for traffic video become increasingly urgent. The newly established HEVC(High Efficiency Video Coding) standard includes traffic video compression as one of its requirements and investigate new coding tools to enhance the coding efficiency of traffic video compression.The latest designed video coding standard called HEVC, high efficiency video coding standard, is successfully published at Jan.2013. The HEVC standard still adopt the compound coding framework and several advanced coding tools are incorporated into the HEVC standard.This paper uses the HEVC standard as an anchor. Designs relevant test platform, followed methods of use reference frame in the interframe encode.And put forward the method of use virtual reference frame to improve the coding efficiency. Mainly divided into the following two aspects:First, considering the characteristics of video images and correlations between images, we propose traffic video encoding scheme base on virtual reference frame, mainly including build the virtual reference frame and manage the virtual reference frame.In order to economizing ues the features such as significant difference between foreground and background and little changing with light, we propose virtual reference frame as a supplement of the inter-frame prediction. The so-called virtual reference frame is builded with blocks or pixels in a kind of rule. Corresponding block matching degree or frequency of the same pixels was used as the key factors to decide whether the blocks or pixels are the constituent part of virtual reference frame, then build the background reference frame use the imagesequence, improve the reference frame management mechanism of RPS(Reference Picture Set) and put the virtualreference frame into reference frame list.Experimental results show that the proposed scheme improve the coding performance significantly compared with HEVC.Second,in order to make full use of the virtual reference frames' features, real time update and stage update of virtual reference frame were proposed as update strategy. When one or more frames were encoded, the virtual reference frame used now should be updated.By means of updating of the virtual reference frame,effectively reduce the affection caused by luminance change and camera angles change.Experimental results show that the proposed scheme about updating the virtual reference frame improve the coding performance, especially the number of video images is more.