Multi-objects colorful segmentation algorithm for complex background

Abstract

The background modeling of multi-object segmentation of moving video relies much on the change of environment. Making use of the background subtraction directly may lead to unsatisfactory even wrong segmentation. An improved algorithm of the codebook background modeling based on Kalman filtering theory was proposed. According to the codebook,a colorful model for each pixel was built to distinguish the foreground and background pixels,and according to the characteristics of temporal recursive low passed of Kalman filtering,it was used to rectify the codebook background modeling. The experimental results show that the improved algorithm can update background model efficiently,has the strong anti-jamming ability and segregate the moving objects accurately to meet the real-time requirements.