Adaptive shadow identification through automatic parameter estimation in video sequences

This paper introduces a new algorithm for the discrimination of moving object and moving shadow pixels in video sequences. The algorithm works in HSV space and is designed to be unaffected by scene type (indoor, outdoor), background type (plain, patterned, complex) or lighting conditions (direct sunlight, fluorescent lights, diffuse lighting). In order to operate successfully in both indoor and outdoor environments (possibly with dynamically changing illumination) without the need for manual calibration, or the manual adjustment of any parameters, an algorithm for the dynamic determination of the parameter required by this method is also presented.