Reflection Components Separation based on Chromaticity and Noise Analysis

Highlight reflected from inhomogeneous objects is the combination of diffuse and specular reflection components. The presence of highlight causes many algorithms in computer vision to produce erroneous results. To resolve this problem, a method to separate diffuse and specular reflection components is required. This paper presents such a method, particularly for objects with uniformly colored surface whose illumination color is known. The method is principally based on the distribution of specular and diffuse points in a two-dimensional chromaticity intensity space. We found that, by utilizing the space, the problem of reflection components separation can be simplified into the problem of identifying diffuse chromaticity. In our analysis, to identify the diffuse chromaticity correctly, an analysis on noise is required, since most real images suffer from it. Unlike existing methods, the proposed method is able to identify diffuse chromaticity robustly for any kind of surface roughness and light direction, without requiring diffuse-specular pixel segmentation. In addition, to enable us to obtain a pure-white specular component, we present a handy normalization technique that does not require approximated linear basis functions.