BP algorithm for background estimation based on temporal and spatial correlation

Abstract

Video background estimation of static fixed-focus camera is the basis for subsequent video analysis and processing. Due to the illumination changes, video background estimation is still an obstacle in video segmentation. A new belief propagation algorithm for video background estimation is analyzed based on the temporal and spatial correlation. Under the condition of different temporal and spatial correlation functions, the corresponding performances of this algorithm are compared. And for the illumination changes, a new temporal correlation function has been proposed. Appending the post processing of gray-scale transformation, the proposed algorithm can effectively deal with the problem of video background estimation under the condition of the rapid illumination changes. The experimental results verify the validity and reliability of this algorithm in several scenarios.