DepthComp: Real-time Depth Image Completion Based on Prior Semantic Scene Segmentation

Atapour-Abarghouei, A.; Breckon, T.P.

Abstract

We address plausible hole filling in depth images in a computationally lightweight methodology that leverages recent advances in semantic scene segmentation. Firstly, we perform such segmentation over a co-registered color image, commonly available from stereo depth sources, and non-parametrically fill missing depth values based on a multipass basis within each semantically labeled scene object. Within this formulation, we identify a bounded set of explicit completion cases in a grammar inspired context that can be performed effectively and efficiently to provide highly plausible localized depth continuity via a case-specific non-parametric completion approach. Results demonstrate that this approach has complexity and efficiency comparable to conventional interpolation techniques but with accuracy analogous to contemporary depth filling approaches. Furthermore, we show it to be capable of fine depth relief completion beyond that of both contemporary approaches in the field and computationally comparable interpolation strategies.

Citation

Atapour-Abarghouei, A., & Breckon, T. (2017, September). DepthComp: Real-time Depth Image Completion Based on Prior Semantic Scene Segmentation. Presented at 28th British Machine Vision Conference (BMVC) 2017, London, UK

• Presentation Conference Type: Conference Paper (published)
• Conference Name: 28th British Machine Vision Conference (BMVC) 2017
• Start Date: Sep 4, 2017
• End Date: Sep 7, 2017
• Acceptance Date: Jul 1, 2017
• Online Publication Date: Sep 4, 2017
• Publication Date: 2017
• Deposit Date: Jul 20, 2017
• Publicly Available Date: Jul 21, 2017
• Pages: 208.1-208.13
• Book Title: Proc. British Machine Vision Conference
• DOI: https://doi.org/10.5244/C.31.58
• Keywords: depth filling, RGB-D, surface relief, hole filling, surface completion, 3D texture, depth completion, depth map, disparity hole filling
• Public URL: https://durham-repository.worktribe.com/output/1145974
• Publisher URL: https://breckon.org/toby/publications/papers/abarghouei17depthcomp.pdf