appliying your method to filling holes of kineck or stereo depth map images !

[RedOne88]

Thank you for sharing your code! I would just like to know the possibility of applying your code on normal depth images, like depth map image of kinect or stereo! the idea is to fill the invalid pixels (black pixels in attached image ) . Thank you in advance !

[TixiaoShan]

@RedOne88 I think this might be the paper you want to use. Learning to Reconstruct 3D Structures for Occupancy Mapping http://www.roboticsproceedings.org/rss13/p09.pdf

[kujason]

While this algorithm isn't specifically designed for that task, the masked morphological operations can be used to fill in the holes (see https://github.com/ethnhe/PVN3D and https://arxiv.org/abs/1808.03833 for examples)