Research On Algorithms Of Image Shadow Detetion And Removal

The shadow in an image is a kind of interference to the vision and image information of computer vision, which can reduce the image quality, and make some processings unstable or failed such as object segmentation, tracking and recognition. In this thesis, we research the algorithms for shadow detection and removal for making the vision and image information processing more robust.Firstly, we introduce how to model light source, object reflection and imaging of human eye or vidicon, and give several color models based on photometry and colorimetry. Forthermore, we analyse the existing conditions and characters of shadows.In the researches on shadow detection, we firstly introduce two algorithms, including the theories and methods of them, which are based on illuminate invariant image and invariant color features respectively, and both of which are based on trichromatic theory. Next, a novel shadow detection algorithm, based on illumination invariance in Gaussian color model is presented at the view of tetrachromatism. The experimental results show the method we present can detect shadows effectively.In the researches on shadow removal, we investigate three algorithms which are based on integral, illumination change surface and texture-consistent respectively. We first introduce the theories and the methods of them and then through comparing the experimental results we analyse the advantages and disadvantages of them.