Using intrinsic images for shadow handling

The ability to detect shadows is a critical feature of any intelligent transportation system (ITS). The improper handling of shadows can be the cause of erroneous conclusions in traffic analysis. As vision-based ITS applications are becoming more popular, it is important to minimize the effects of shadows. Here, we present a novel method for shadow handling in a sequence of images. Following recent work motivated by studies of the statistics of natural images, we show that this new method is a reliable detector of shadows and can be easily implemented in real-time.