Illumination Component Separation Method Based on Independent Component Analysis

In the field of computer vision, since the image of an object is negatively influenced by illumination, material, texture, shadows, and so on, it is often necessary to separate the illumination components (diffuse and specular reflections) from the intrinsic information of an image. In this paper, the diffuse and specular reflections are regarded as mutually independent components. By analyzing the coefficient matrix of the illumination components and incorporating it into independent component analysis, the diffuse and specular reflections are separated. The experiment results reveal that the proposed method is feasible and has a better separation precision.