An Adaptive Shadow Removal Method Based on Weighted Gaussian Model

The method of the shadow detection and remove which has application value is proposed based on analyzing visual surveillance purposes. The method proposed is essentially based on linear ratio and shadow direction angle threshold to locate the rough shadow regions, considering the relationship of the probability whether it is a shadow pixels and distances which between the shadow pixels and the centroids of the foreground objects. So this paper use the distances to weight the probability regions, and then exactly remove the shadow using the threshold method to locate the exact shadow regions. The experimental result indicates that the algorithm of removing shadow is fast and exact, improves the accuracy of the moving vehicle detection.