Abstract: To estimate dynamic changing illumination of outdoor scenes,a real-time video based illumination estimation algorithm is proposed.It adheres to the linear decomposition model of outdoor scene image and fully explores the lighting features of shadowed pixels as well as the illumination coherence between adjacent frames to construct an optimization function to solve for the incident intensity of skylight and sunlight.Experimental results demonstrate the robustness of the proposed algorithm as well as its applicability in augmented reality,shadow detection and relighting.