Research On Based On BP Algorithm For Background Estimation

Video background estimation is to reconstruct the static background image from the image sequence of video which is from the static fixed-focus camera and containing the moving object; which is to solve segmentation of the static and moving object from the video image sequence and the basis for subsequent video analysis and processing; the reliability and accuracy of the results is essential for the entire video information processing system; so the issue has become a hot issue of video information processing. Because of the actual environment in the relativity of movement and stillness, the background illumination changes by factors such as the video background variability and background of long dense moving target lead to the shelter, making it difficult to accurately estimate the background. This paper focuses on the video background estimation, and optimizing the computational efficiency of belief propagation algorithm.In the algorithm of video background estimation base on the belief propagation algorithm; according to the video background spatial and temporal correlation, analysis and comparison of each kind of existing spatial correlation functions that influence estimating background. On temporal domain, simulating delta modulation we has defined a new temporal domain correlation function, which could improve the accuracy of estimating the video background. And applied the corresponding methods to the moving object detection, which confirmed the method is validity. In the last, we have given the simulation results of AIRPORT and PETS2001 image sequences.According to the local correlation distribution of the reference background, using the local maximum weight match to optimizing the path of information transfer, thus make full use of effective information in the iterative process, to achieve optimizing the operational efficiency of the belief propagation algorithm. The simulation result of AIRPORT and PETS2001 image sequences proved the algorithm not only improved the operational efficiency, but also improved the accuracy of the video background estimation.The video background estimation of basing on the belief propagation algorithm could rapidly reconstruct the background in complicated environment such as illumination change and leaves crembling, improve the computational efficiency and effective to avoid appearing over-smoothing problem in the video background estimation by optimizing the path of information transfer.