Texture-based shadow removal from a single color image

Abstract

Shadows can often be found in images and videos acquired under strong directional lighting, such as those taken outdoors on a sunny day. Shadows can cause many difficulties in image processing and vision-related tasks, such as image segmentation and object recognition. Therefore, shadow removal is needed for improving the performance of these image understanding tasks. We present a new shadow removal algorithm for a singe texture color image. Intra-image textures are popular and frequently influenced by shadow. Through intra-image texture matching and deconvolution on derivative maps of original image with zero values set in shadow edge positions, we can recover shadow free images. The experimental results on real-world data show good performance of this algorithm. © 2008 Springer Berlin Heidelberg.