Depth Inpainting database
This dataset contains depth images and masks for depth image inpainting.
Dataset in zip:
The description of the files provided in the dataset:
This dataset contains 16 subdirectories. The subdirectories contain depth images in png format converted from the ground truth disparity maps of the Middlebury Stereo Dataset . In each subdirectory, disp.png is the ground truth disparity map (pixel ranges from 0 to 255). Unknown values of the Middlebury dataset are converted to 0s in disp.png.
Accompanied with the depth images are the missing masks (mask.png or mask_*.png) and the damaged images (missing.png or missing_*.png).
Other images are corresponding inpainting results of the algorithms mentioned in our paper. See more details in our paper.
| original depth
|| missing depth
| inpainted by LR
|| inpainted LRL0
|| inpainted by LRL0^\psi
If you used the processed data sets on this page, we appreciate it very much if you can cite our following works:
Hongyang Xue, Shengming Zhang, Deng Cai: Depth Image Inpainting: Improving Low Rank Matrix Completion with Low Gradient Regularization, IEEE Trans. Image Processing 26(9): 4311-4320 (2017).
Our dataset is made based on the Middlebury Stereo Dataset:
||D. Scharstein, H. Hirschmüller, Y. Kitajima, G. Krathwohl,
N. Nesic, X. Wang, and P. Westling.
High-resolution stereo datasets with subpixel-accurate
ground truth. (GCPR 2014),
Return to Codes and Data