Research On Object Clarity Method In Video Surveillance

With the development of digital images, video surveillance systems are widelyused in daily life, and become an important part of the security system. However, theimage will become blurry because of the inevitable effect of the hardware and theexternal environment during the procedure of acquisition, transmission and storage,resulting in the decline in the overall quality of the image, thus affecting the follow-upworks such as the target tracking and recognition. Fuzzy image sequences are obtainedfrom the video surveillance system first, and the moving objects are detected andextracted. Based on the above procedure, using clarity algorithms restore the extractedobjects and more clear objects are got. The main work of the paper is as follows:Before the research of clarity algorithms, the moving objects are detected andextracted from the video surveillance images in order to reduce computation andimprove the system in real-time. After the description and analysis of several classicmethods of moving objects detection, an improved moving objects detection algorithmis proposed, which considers the advantages and disadvantages of the temporaldifferencing method and the background subtraction method. The algorithm is based onthe dynamic threshold image, and is accomplished through the temporal differencingand background subtraction with the background updated in real time. So the moreaccurate moving objects are obtained. In the research of the object clarity in videosurveillance, the reasons causing blurred images in the video surveillance system areanalyzed, and the clarity algorithms are studied through the image enhancement andrestoration respectively. The Wiener filter, constrained least squares method andhomomorphic filtering are used in the surveillance image to get the clear ones.Meantime a new image restoration algorithm is proposed which is based on theestimation of the information of the degraded image, that is, the unknown distributionwhich the actual image is blurred by is estimated by the chi-square test. Finally, theprior extracted objects are tested through the Clarity algorithms above, only a specificarea of the image is processed to improve the system in real time.On the basis of these studies, the algorithm in the paper is implemented viaOpenCV and MATLAB, and a clarity processing system of the object in video surveillance is designed and developed on VC++platform. Experiments have beenconducted using collected video files, and the results proved that the clarity algorithmproposed in the paper can process the moving object in the video surveillance effectivewith certain robustness.