Shadow processing with color model adjustment and texture analysis

Object detection is a fundamental task in computer vision. In most cases, this motive is often corrupted by the shadows in an image. These scenarios consequence a great need of shadow processing. Along with this, the method of detecting and removing shadow is used to improve computer vision applications such as image segmentation, object recognition and tracking. The prime objective of this paper is to detect and remove shadow from an image by analyzing color models and background texture pattern. Initially, shadow boundaries are detected from a given foreground region by adjusting color models. Then the similarity between texture features of shadow and neighboring non-shadow region is measured. Finally, based on these similarities, texture pattern of non-shadow region is projected onto the shadow region to get a shadow free image. However, it is noteworthy that Local Binary Pattern (LBP) is used here to measure the texture feature as it is simple and efficient. In addition, this simple methodology has achieved a good detection rate of 87.81 % and presented a high PSNR (22.41), SSIM (0.9432) value and low RMSE (3.48) value after shadow removal.