Incremental-LDI for multi-view coding

Abstract

This paper describes an Incremental algorithm for Layer Depth Image construction (I-LDI) from multi-view plus depth data sets. A solution to sampling artifacts is proposed, based on pixel interpolation (inpainting) restricted to isolated unknown pixels. A solution to ghosting artifacts is also proposed, based on a depth discontinuity detection, followed by a local foreground / background classification. We propose a formulation of warping equations which reduces time consumption, specifically for LDI warping. Tests on Breakdancers and Ballet MVD data sets show that extra layers in I-LDI contain only 10% of first layer pixels, compared to 50% for LDI. I-LDI Layers are also more compact, with a less spread pixel distribution, and thus easier to compress than LDI Visual rendering is of similar quality with I-LDI and LDI.