Moving Shadows Removal using HSV Color Space and Texture Analysis

The paper presents a new approach for detection of moving shadows. The approach is based on the assumption that shadow regions are darker than the corresponding background but have the same chromacity and texture. We use both HSV and RGB color spaces to extract spectral information and combine two texture features to detect moving cast shadows. Firstly, candidate shadow regions are extracted by spectral analysis. Secondly, we use proposed extended Local Ternary Pattern in the paper, improved edge information, and HSV color space analysis to attain three shadow detectors which are used to vote to select the final shadows, respectively. Our approach is much more effective than other five typical algorithms according to the experimental results.