适于交通监控的内容自适应空间可伸缩视频编码算法  

CONTENT-ADAPTIVE SPATIALLY SCALABLE VIDEO CODING ALGORITHM SUITABLE FOR TRAFFIC SURVEILLANCE

作　　者：时秀波[1] 

机构地区：[1]威海职业学院信息工程系,山东威海264210

出　　处：《计算机应用与软件》2015年第7期323-327,333,共6页Computer Applications and Software

摘　　要：当前空间可伸缩编码很少考虑视频感兴趣区域ROI(Region of Interest)或视觉突出内容,从而不能更好地适应视觉重要内容在较低分辨率移动终端的显示。对此,提出一种适用于交通监控的内容自适应空间可伸缩视频编码算法。采用背景差法获取运动车辆并进行目标跟踪,一般情况下在交通监控中主导车辆(视觉上最突出的车辆)所在运动窗为ROI,将该ROI设定为裁剪窗口,并使用H.264/AVC可伸缩编码SVC(Scalable Video Coding)标准的扩展空间可伸缩方法 ESS(Extended Spatial Scalability)进行编码,同时在空间增强层使用跟踪准确度代替PSNR作为质量度量标准。实验表明,该算法与传统的下采样空间扩展算法相比,对低分辨率空间层解码下的视觉感知有较大的改善,而且在保证跟踪相对准确的前提下,码率可以节省约60%。Regions of interest （ROI） or visually salient contents are rarely considered in current spatially scalable video coding, thus the display of visually important content can not be better adapted to the mobile terminals with lower resolution. In light of this, we propose a content-adaptive spatially scalable coding algorithm suitable ＂for traffic surveillance video. It uses background subtraction algorithm to detect moving vehicles and track targets, generally speaking the motion window the leading vehicle （the most visually protrudent vehicle） in traffic surveillance located is referred to as ROI, and the ROI is set as the cropping window and is encoded with the extended spatial scalability （ESS） of H. 264/AVC scalable video coding （SVC） standard. Meanwhile, in space enhancement layer the tracking accuracy is used instead of PSNR as the quality metric criterion. Experiment shows that compared with conventional downsampling spatial scaling coding algorithm, the proposed algorithm can greatly improve the visual perception in decoded low-resolution space layer, and allow for about 60% bit rate savings while maintaining comparable tracking accuracy.

关 键 词：可伸缩视频编码 扩展空间可伸缩性 交通视频 内容自适应 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]