监控系统中的对象分割与视频压缩研究

项目名称： 监控系统中的对象分割与视频压缩研究

项目编号： No.61001082

项目类型： 青年科学基金项目

立项/批准年度： 2011

项目学科： 金属学与金属工艺

项目作者： 刘雪冬

作者单位： 武汉理工大学

项目金额： 7万元

中文摘要： 随着视频监控系统的广泛应用，相关的智能化技术成为研究热点，而针对监控视频的压缩研究相对薄弱。本项目基于监控场合提供的便利条件针对监控视频的特殊要求对监控视频压缩技术展开研究。监控对象的实时分割对于基于对象的视频编码尤为重要，本项目建立了全景相机和监控摄像机成像的几何模型，导出了视频帧像素和全景图像素位置之间的理论关系，实现了利用高分辨率全景图象和监控摄像机运动参数进行视频背景提取。在编码方面，摒弃了传统的PSNR 客观准则，先对监控对象进行可控增强以改善其主观质量，然后再进行高保真度压缩，同时提高了编码效率和速度。

中文关键词： 全景图像；摄像机运动参数；可控图象增强；高保真度压缩

英文摘要： With the wide use of video surveillance systems, intelligence-related technologies become a research hotspot. However, research on the surveillance video compression is relativly weak. In this project, based on the convenience provided by surveillance circumstances, surveillance video compression is studied to fulfill the special demand on the surveillance video. Real-time segmentation of surveillance objects plays a very important role for the object-based video coding. An imaging geometry model of the panorama and surveillance cameras is established and a theoretical relation between pixel locations of frame backgrounds and ones of the panorama image, and thus video background extraction can conduct using surveillance camera parametrs along with a high resolution background panorama. For the video coding, the traditional objective PSNR criterion is abandoned and surveillance objects are firstly enhanced to improve the subjective quality, afterwards compression with high fidelity is conducted, and thus the coding efficiency and speed are improved.

英文关键词： panorama image; camera motion parameters; controllable image enhancement; compression with high fidelity