Separating Reflection Components of Textured Surfaces

 Robby T. Tan         Katsushi Ikeuchi

[ PDF | Code ] Abstract: In inhomogeneous objects, a highlight is a linear combination of specular and diffuse reflection components. A number of methods have been proposed to separate these two components. To our knowledge, all methods that use a single input image require explicit color segmentation to deal with multicolored surfaces. Unfortunately, for complex textured images, current color segmentation algorithms are still problematic to segment correctly. Consequently, a method without explicit color segmentation becomes indispensable, and this paper presents such a method. The method is based solely on colors, particularly chromaticity, without requiring any geometrical information. One of the basic ideas is to iteratively compare the intensity logarithmic differentiation of an input image and its specular-free image. Specular-free image is an image that has exactly the same geometrical profile as  the diffuse component of the input image, and that can be generated by shifting each pixel's intensity and maximum chromaticity non-linearly. Unlike existing methods using a single image, all processes in the proposed method are done locally, involving a maximum of only two neighboring pixels. This local operation is useful for  handle textured objects with complex multicolored scenes. Our evaluations by comparison with the results of polarizing filters show the effectiveness of the proposed method. "Do not use the images in this website for testing your code. The images are compressed images whose brightness might not be linear to the flux of incoming light.” Results: input specular free diffuse component specular component input specular free diffuse component specular component I. SFU (SimonFraserUniversity) Database: (note, in the experiments we ignored saturated pixels) 1. Plastic-2_solux-3500: input specular free diffuse component specular component 2. Apples_syl-50MR16Q: input specular free diffuse component specular component II. Examples of Error Separation: 1. Illumination chromaticity is wrongly estimated:     actual illumination chromatcitiy: {r,g}={0.31, 0.31},  we set the illumination {0.35, 0.31}:     input specular free diffuse component (error illumination estimation) specular component 2. Near achromatic surface (parts of melon's gray color is deemed as specularity, resulting wrong separation) input specular free diffuse component specular component [ Abstract ] http://www.staff.science.uu.nl/~tan00109/textureSeparation/results.html