Specularity removal in images and videos: a PDE approach

摘要: We present a unified framework for separating specular and diffuse reflection components in images videos of textured scenes. This can be used specularity removal independently processing, filtering, recombining the two components. Beginning with partial separation provided by an illumination-dependent color space, challenge is to complete using spatio-temporal information. accomplished evolving differential equation (PDE) that iteratively erodes component at each pixel. A family PDEs appropriate differing image sources (still vs. videos), prior information (e.g., highly lightly scenes), or computations optical flow) introduced. In contrast many other methods, explicit segmentation and/or manual intervention are not required. results on high-quality video acquired laboratory addition taken from Internet. Results latter demonstrate robustness low dynamic range, JPEG artifacts, lack knowledge illuminant color. Empirical comparison physical specularities polarization provided. Finally, application termed dichromatic editing presented which processed produce variety visual effects.