Depth Map Refinement for Immersive Video

In this article, we propose a depth map refinement method that increases the quality of immersive video. The proposal highly enhances inter-view consistency maps (estimated or acquired by any method), crucial for achieving required fidelity virtual view synthesis process. described method, only information from is used, as use texture can introduce errors in refinement, mostly due to color inconsistencies and noise. order evaluate performance compare it with state art, three experiments were conducted. To test influence on encoding video, four sets (original, refined synthesis-based bilateral filter, proposal) encoded MPEG Immersive Video (MIV) encoder. second experiment, provide direct evaluation accuracy maps, Middlebury database comparison was performed. third temporal assessed measuring efficiency views. showed both high increase video applications higher similarity ground-truth after estimated maps. usefulness appreciated confirmed experts ISO/IEC group became Reference Software refinement. implementation publicly available other researchers.