A PDE Approach to Specularity Removal in Images Comp 766: Shape Analysis in Computer Vision

Specular reflections are exhibited by a wide range of materials whose reflectance can be described as a linear combination of specular and diffuse components [5]. There are several benefits to separating an image into the two components. By isolating the diffuse component (which is often well-described by the Lambertian model), powerful Lambertian-based tools for tracking, reconstruction and recognition (e.g. shape-fromshading) can be more widely applied to real-world, non-Lambertian scenes. Specular reflectance itself plays an evident role in human perception. Based on this, several computer vision algorithms have been designed to successfully infer shape solely from specularities [3].