Adaptive shadow estimator for removing shadow of moving object

We propose an adaptive shadow estimator to detect and eliminate the shadow of a moving object while adapting to variation of illumination and the environment in an automatic manner. The proposed method discriminates between the shadow and the moving object by cascading three estimators which use the properties of chromaticity, brightness, and local intensity ratio. In the spatial adjustment step, the method compensates for accumulated errors in the cascading process. Experimental results show that our scheme can operate in real-time, outperforms existing methods, and rapidly adapts to variations in the environment.