基于角点特征检测的视频图像质量评价方法  

Video Image Quality Assessment Method Based on Corner Feature Detection

作　　者：周正[1] 田昕[2] 周城[3] 冯伟东[1] 李涛[4] 熊承义[3] 

机构地区：[1]湖北省电力公司信息通信分公司,武汉430077 [2]武汉大学电子信息学院,武汉430072 [3]中南民族大学电子信息工程学院,武汉430074 [4]华中科技大学多谱图像信息处理技术国防科技重点实验室,武汉430074

出　　处：《计算机工程》2013年第9期30-33,44,共5页Computer Engineering

基　　金：国家自然科学基金资助项目(61102064;61201268);智能无线通信湖北省重点实验室开放基金资助项目(IWC2012003)

摘　　要：为对视频会议系统中图像的降质程度进行有效评估,提出一种基于角点特征检测的视频图像质量评价方法。模拟测试图像在真实环境下的降质情况,生成仿真图像。计算两者的二值边缘轮廓曲率,获得不同检测参数下的角点检测数,寻找用于真实环境的最佳测试图像和角点检测参数。根据角点检测数将图像质量分成若干个等级,并依此判别图像质量的好坏。Matlab仿真实验结果表明,该方法得到的结果与主观评价结果一致,能正确评价图像质量。In order to assess the image degraded quality problem of a video, an objective video quality assessment method based on comer feature detection is proposed. By simulating the test images degraded in a real environment, and simulated images are generated. Then the number of corner in different corner detection parameters for test images and simulated images are computed by binary edge image to curvature. The best test image and corner detection parameters for a real environment can be obtained. Furthermore, the video quality is divided into several grades by the number of the detected corner. Then the number of corner for the test image detected in a real environment is regarded as the discriminant basis of the video quality assessment. Simulations are carried out by Matlab, the results show that the result of proposed method is keeping with subjective evaluation results, it is efficient.

关 键 词：角点特征检测 视频图像 图像质量评价 曲率尺度空间 CANNY算子 

分 类 号：TN919.8[电子电信—通信与信息系统]