PROJECT TITLE :

Rate-Distortion Driven Decomposition of Multiview Imagery to Diffuse and Specular Components

ABSTRACT:

We provide a method for compressing multiview photography that utilises an overly full representation. Multiview datasets are decomposed into two additive portions, diffuse and specular content, using a rate-distortion (R-D) driven technique. Decomposition is driven exclusively by compressibility by employing an R-D-inspired measure as our optimization cost function for the diffuse and specular components. For the sake of simplicity, we begin by outlining a framework for doing data separation within a registered domain. Specular data can then be separated from the coordinates of various reference views using a more complete approach. A coding boost of up to 0.6 dB for synthetic datasets and up to 0.9 dB for real datasets has been demonstrated in experiments.