Separating Specular and Diffuse Reflection Components in the HSI Color Space

In this paper, we propose to separate diffuse and specular reflection components for color images in the HSI color space. Under white illumination, pixels with the same diffuse chromaticity have the same hue. Meanwhile, specular pixels have lower saturations than the diffuse ones. Based on these properties, separating reflection components can be achieved by adjusting saturations of specular pixels to the values of diffuse-only pixels with the same diffuse chromaticity. We employ a region-growing algorithm to locate adjacent pixels with similar diffuse chromaticities. Then, the separation of reflection components is achieved byfinding the optimal saturation in each connected region. The experimental results demonstrate that the proposed method is more effective to separate reflection components than the state-of-the-art methods.