Background update algorithm based on blocks classification for intelligent video surveillance
Lizhong Xu · Caai Transactions on Intelligent Systems · 2010
Background update algorithms have excessive calculation overhead and are sensitive to changes in light-ing.In order to solve these problems,a background update algorithm based on block classification was proposed.First,image differences were obtained by subtracting the incoming frame from the reference image.Then the image differences were divided into blocks of equal size.Each block was then classified as a background block or a fore-ground block according to the blocks’predominant features.Different updating strategies were then employed ac-cording to the classification of the block.In this way,real-time background updates were possible.This algorithm overcame problems of computational redundancy arising in other pixel-background models.Execution speed was im-proved because object-operations were performed on every block.Experimental results showed that this method well adapts to changes in illumination.