Moving cast shadow detection based on regional growth

Moving cast shadow detection is important in the detection and tracking of moving objects. This paper proposes an approach based on regional growth to detect moving cast shadow. Firstly, the pixel distribution histogram in RGB color space or the luminance ratios in HSV color space is used to detect the possible shadow area, which can produce a possible shadow area to reduce the calculation of subsequent processing. Secondly, we implement the regional growth approach based on the Breadth-First Search algorithm to get a relatively accurate shadow area. Our approach considers both the color information and the edge features of images, which yields accurate detection of moving cast shadows as shown by experiments.