Background reconstruction using DWT and grayscale classification

Abstract

Background reconstruction is very important in many video-based tracking systems. The principle difficulties are the quality and velocity of reconstruction. To cope with these problems, a novel method is proposed. Firstly, the sequence images are decomposed into low frequency subimages using DWT (discrete wavelet transform). Then, the improved grayscale classification is introduced to reconstruct initial background with the latest N frame sub-images. Finally, the background is updated with selective update and background adjustment. Since sub-images are with lowresolution, the reconstruction cost is decreased. With the accumulation sum being introduced to classify grayscales, the background noise is reduced. The experimental results show that the proposed method is efficient. © 2010 IEEE.