Identification of topological measures of visibility graphs for analyzing transitions in complex time series

Abstract

We present a super-resolution technique where the 3-D shape preservation is used as a constraint while super-resolving a scene. Given the observations under different illuminant positions, we combine these observations to obtain the super-resolved image and the spatially enhanced scene structure simultaneously. The use of shape cue in the form of photometric measurements, instead of the motion cue, eliminates the need for image registration with a sub-pixel accuracy. We model the high-resolution image, the structure, and the albedo of the surface as separate Markov random fields and super-resolve them using a suitable regularization scheme. Results of the experimentation on real data are presented.