Background model of complex scene based on level information synthesis

A level information synthesis model(LISM)is proposed to subtract and maintain background for practical visual traffic surveillance applications.LISM is composed of three components:the pixel level component that performs background candidate classification by analyzing the pixel state;the region level component that recognizes regions of background from background candidate areas by the rotation-invariant enhanced local binary pattern(RELBP)texture analysis;and the frame level component that detects and copes with global changes in the background,especially uses the revised histogram modification(HM)to handle sudden change of environmental illumination.Finally,the experiments results proved the feasibility of the LISM algorithm.