监控视频图像质量诊断方法研究及实现

发布时间：2018-02-27 20:51

本文关键词： 监控视频图像质量检测偏色清晰度　出处：《华东理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：视频监控系统是安全防范系统的重要组成部分,随着智能监控的发展,监控系统的范围急剧扩大,系统中监视器的数目也随之增加,这都加重了监控系统的维护检测难度。由于智能监控对图像质量有一定的要求,如果图像质量异常而不能及时发现并解决,很可能造成智能监控部分功能无法正常工作,进而造成一定的经济损失,人工检测费时费力,工作量大,不利于降低成本。实现视频监控系统的自动检测,将有效降低人工检测的压力,也可起到降低成本的目的。本文针对监控视频系统中极易出现的图像帧异常情况,构建了基于目标检测的视频图像质量客观诊断方法,对视频图像中是否存在雪花干扰、滚屏、信号丢失、画面冻结、丢帧、偏色和图像模糊等异常现象进行检测,实现了图像质量客观诊断方法各异常项的检测,并对其进行模块封装。在视频输入之后,工作人员可以通过本文设计的诊断模块对图像进行质量检测,提高故障排除的效率,降低人工检测成本。本文提出了基于a-b分布最大轮廓重心距离的偏色检测方法和基于结构相似度的图像清晰度评价方法,与传统方法进行了对比,取得了较好的效果。测试结果满足人眼的主观评价要求。
[Abstract]:Video surveillance system is an important part of security prevention system. With the development of intelligent monitoring, the scope of monitoring system expands rapidly, and the number of monitors in the system also increases. This makes it more difficult to maintain and detect the monitoring system. Since intelligent monitoring has certain requirements for image quality, if the image quality is abnormal and can not be detected and solved in time, it is likely that the functions of the intelligent monitoring part will not work properly. This will result in certain economic losses, time-consuming and laborious manual detection, which is not conducive to reducing the cost. The realization of automatic detection of video surveillance system will effectively reduce the pressure of manual detection, but also play the purpose of reducing the cost. In this paper, the objective diagnosis method of video image quality based on target detection is constructed, which is aimed at the abnormal situation of image frame that is easy to appear in the surveillance video system. Whether there is snowflake interference, scrolling screen, signal loss, picture freezing in the video image or not, this paper proposes an objective diagnosis method based on target detection. The abnormal phenomena such as frame loss, color deviation and image blur are detected, and the detection of the abnormal items in the objective diagnosis method of image quality is realized, and the module is encapsulated. After the video input, The staff can use the diagnostic module designed in this paper to detect the image quality and improve the efficiency of troubleshooting. In order to reduce the cost of manual detection, this paper presents a new method of color deviation detection based on the maximum centroid distance of a-b distribution and an image definition evaluation method based on structural similarity, which is compared with the traditional method. The test results meet the subjective evaluation requirements of human eyes.
【学位授予单位】：华东理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN948.6

【参考文献】