Separation of Diffuse and Specular Reflection Components from Real-world Color Images captured under Flash imaging conditions

Abstract

Image processing applications like endoscopy, flash photography are prone to specularity reflection because of the usage of flash illumination during image acquisition. The formation of specularity reflection paralyzes the performance accuracy of many state-of-the-art feature detection algorithms. Specularity removal in case of realworld imaging conditions is further challenging when the source of illumination is unavailable. Literature suggests that chromaticity based reflection removal algorithms circumvent the illumination source dependency by transforming the pixels into the chromaticity-intensity plane and solving a fine-grained, least-squares problem of the dichromatic model. In this paper, we extend the existing chromaticity based removal approach to HueSaturation-Value color domain and explore its suitability in specularity removal under flash imaging conditions. Experimentation on MIT intrinsic database demonstrate that our approach achieves desirable reflection separation results with minimum execution time compared to the state of the art.