Study On Shadow Detection And Removal

The presence of shadows in images can lead to problems of some computer vision tasks such as image segmentation, object recognition and tracking. This thesis aims at doing some research on shadow detection and removal algorithms in still images.Shadow detection is the first step of image shadow processing. The result of shadow detection will directly affect the following shadow removal algorithm. We summarize existing shadow detection algorithms and analyze the advantages and disadvantages of them. Based on recent studies, that the red-green channel of opponent color space is invariant to shadows, we propose a method which using the two chromaticity channels of opponent color space to identify shadows boundaries. Experiments validate our approach.For the second part of this thesis, a novel approach of shadow removal is presented. The method is based on the theory of color constancy. The original image is first divided into shadow region and non-shadow region. After that, color constancy algorithms are used to estimate the illuminant color of both regions, the scaling constant between them is calculated. Finally, we get the shadow free image by adding the scaling constant to shadow region pixels. Experimental results show that the method succeeds in removing shadows and is simpler and faster than state-of-the-art methods.