Specularity Removal using Dark Channel Prior

The reflectance of inhomogeneous objects can be described as a linear combination of diffuse and specular reflection components. Most computer vision algorithms assume that visually observable surfaces consist only of diffuse reflection. The existence of specular reflection can be misleading to these computer vision algorithms. A new algorithm  dark channel prior based specularity removal is proposed for separating specular and diffuse reflection components on colorful surfaces from a single input image. The dark channel prior is applied to detect the specular pixels in the image. The maximum diffuse chromaticity of the diffuse pixels is propagated to their neighboring specular pixels after specularity have been detected. Specularity removal can be achieved by using the specular-to-diffuse mechanism. The experimental results show that the proposed algorithm obtain comparable results as the state-of-the-art reflection components separation methods with the merit of being computationally more efficient.