Real-Time Specularity Detection Using Unnormalized Wiener Entropy

Specularity is a common phenomenon in the real world and can provide both challenges and opportunities for computer vision algorithms. In this paper, we propose a new method to detect specularities in real time. This method is based on an unnormalized version of the Wiener Entropy which is commonly used in audio spectral analysis. Experiment results demonstrate our proposed methodology's efficacy in specularity detection on both natural and synthetic images. Its potential in the 3D movie industry and for helping compute stereo correspondence in the presence of specularities is described.