Shadow removal from uniform-textured images using iterative thresholding of shearlet coefficients

摘要

Shadows are natural phenomena that appear in images due to inconsistent illumination of the scene being captured. Recently, the need for removal of shadows from images and videos has gained wide attention due to the ill-effects of shadows on many computer vision tasks. This paper presents a novel technique to remove shadows from images with a uniform background. Initially, our method identifies the shadow and the lit regions by discarding the low-frequency image details. This is followed by an iterative procedure in which the shadow pixels to be corrected are located by eliminating the Shearlet approximation coefficients greater than a threshold. The shadow pixels identified in each iteration are corrected using a pre-computed correction factor. The shadow-corrected image is finally inpainted to generate the shadow-free output. In order to demonstrate the superior performance of the proposed method, we provide both qualitative and quantitative comparisons of the method with other state-of-the-art techniques.