A fast background estimation method for vehicle surveillance

This paper presents a fast background estimation method for vehicle surveillance based on multi-resolution analysis (MRA) and block updating strategy. Firstly, an approximate expression of original image is achieved by (MRA) and the feature difference is calculated along adjacent frames to divide the foreground and background; Secondly, the isolated pixels are removed by morphologic operation; and then each block of the original image is classified to foreground or background according to the relationship of the multi-resolution analysis; finally, the background is updated block by block based on the hypotheses that moving object is not fixed in a certain place. The proposed method realizes the estimation and updating of background by block and the application of MRA can improve the computation efficiency. Experimental results show the validity and the blur effect of the conventional methods is avoided.