A Recovery Method for Kinect-Like Depth Map Based on Color Image Segmentation

The original depth map captured by low-cost depth cameras such as Kinect suffers from problems of depth missing and noise, which restrict its further application. In this paper, we propose a novel recovery method of Kinect-like depth map utilizing the corresponding color image. The proposed method consists of two steps: hole-filling by image in painting and denoising by fast guided filter. Experiments show that our method gives satisfactory result.

[1]  Hui Chen,et al.  Virtual View Synthesis Based on DIBR and Image Inpainting , 2013, PSIVT.

[2]  Jian Sun,et al.  Fast Guided Filter , 2015, ArXiv.

[3]  Xiaojin Gong,et al.  Guided inpainting and filtering for Kinect depth maps , 2012, Proceedings of the 21st International Conference on Pattern Recognition (ICPR2012).

[4]  Yao Wang,et al.  Color-Guided Depth Recovery From RGB-D Data Using an Adaptive Autoregressive Model , 2014, IEEE Transactions on Image Processing.

[5]  Maziar Loghman,et al.  Refinement of depth maps generated by low-cost depth sensors , 2012, 2012 International SoC Design Conference (ISOCC).