A comparative review of plausible hole filling strategies in the context of scene depth image completion

Abstract

The full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, for personal research or study, educational, or not-for-prot purposes provided that: • a full bibliographic reference is made to the original source • a link is made to the metadata record in DRO • the full-text is not changed in any way The full-text must not be sold in any format or medium without the formal permission of the copyright holders.

Real-Time Monocular Depth Estimation Using Synthetic Data with Domain Adaptation via Image Style Transfer

“…Moreover, despite the fact that depth holes are generally considered undesirable [3,4,2,49,58], certain areas within the scene depth should remain without depth val- Figure 8: Results on the Make3D test set [66]. Note the quality of our outputs despite the vast differences between this dataset and the images used in our training.…”

Motion parallax for 360° RGBD video

“…We propose here a depth refinement technique that is inspired by these works but tailored for our end goal. For a more in-depth review of related techniques, we refer the reader to a recent survey on hole filling strategies for depth map completion [2].…”

Monocular Segment-Wise Depth: Monocular Depth Estimation Based on a Semantic Segmentation Prior

“…As 3D scene understanding is gaining increasing significance within computer vision applications, accurate and efficient depth estimation is now an integral part of many such systems. While strategies such as stereo correspondence [1], structure from motion [2] and depth from shading and light diffusion [3,4] have produced promising results, prevalent issues such as missing values (holes), depth inhomogeneity and computationally intensive processing or calibration requirements are ubiquitous within such approaches [5]. This has given rise to the necessity of depth refinement post estimation [6,7,8].…”