Deriving reflectance and shading components from a single image

Abstract

Interpreting real-world images require the ability distinguish the different characteristics of the scene. Reflectance images are invariant under different illumination conditions, they are more appropriate for some applications, such as object detection, object recognition. Shading images can be used for some research fields such as the analysis of scene illumination and color constancy. Reflectance and shading components of image are represented by intrinsic images. In this paper, for processing the images including the specular reflection and a variety of shadows, an algorithm is presented that uses multiple cues to recover reflectance and shading image from a single image. On the gradient domain of logarithmic image, firstly, a method based on c <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> c <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> and o <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> o <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> color space, called weighted edges map method, is used to design gradient classifier, executing intrinsic image preliminary decomposition. Then intensity characteristics and chromatic characteristics are used to compensate gradient classifier. Experimental results show that the recovering reflectance image by the proposed method can eliminate highlights and shadows effectively and keep more complete color and edge information.