Research And Implementation Of Video Quality Diagnostic Algorithm

With the development of the computer vision technology, the digital image processing and the electronic technology, the intelligent visual surveillance system has become a more important safe defense way. However, with the video surveillance points growing, using human eyes troubleshoot the fault is not only labor-intensive, but also inefficient, delayed and unreliable; Whether it is video surveillance or video analysis, firstly it must ensure that the video transmission over the camera is correct, if the video signal of the camera is interference or damage, the video monitoring and analysis based on content would be meaningless. Therefore, diagnosis of video quality for intelligent monitoring timely and efficiently is of great significance.In this paper, firstly, has a brief overview of the video quality diagnostic studies, and then extract three common factors that influenced the video quality:image noise, image definition, camera interference. Finally, design and implementation of video quality diagnostic algorithms aim at the three main factors.In image noise detection, according to the information of this pixel and around the point, we propose an image noise detection algorithm based on direction operator. The preliminary judging noise points according to eight different directions operators and then using local fuzzy membership function to judge the noise points. The experimental results show that the method is accurate to a large extent.About the detection problem of image definition, an image definition detection algorithm has been proposed based on the width of the transition edge. Using the knowledge of edge information and probability, I gave the definition of the image detection model and test the algorithm model. The results show that the proposed algorithm can detect the image clarity and have certain robustness.A new camera interference detection algorithm has been proposed, it different from existing algorithms, the proposed algorithm can distinguish shift and occlusion, feature detection and matching are applied to the camera interference detection. Use the knowledge of statistical analysis to detect camera shift and occlusion. Experimental results show that the method has highly detection accuracy, lowly false detection and not change with light and weather.