Polarization Guided Specular Reflection Separation

Since specular reflection often exists in the real captured images and causes deviation between recorded color intrinsic color, separation can bring advantages to multiple applications that require consistent object surface appearance. However, due of an is significantly influenced by illumination, existing researches still suffer from near-duplicate challenge, is, becomes unstable when illumination close color. In this paper, we derive a polarization guided model incorporate information into designed iteration optimization strategy separate reflection. Based on analysis polarization, propose generate chromaticity image, which able reveal geometrical profile input image complex scenarios, e.g. , diversity illumination. The accurately cluster pixels with similar diffuse We further use all these clusters as implicit prior ensure component will not be mistakenly separated component. With model, reformulate unified function solved ADMM strategy. detected jointly RGB polarimetric information. Both qualitative quantitative experimental results have shown our method faithfully reflection, especially some challenging scenarios.