a small question

[hucaigang]

When i was reading your paper,i found the experiments conducted using some sturcture missing depth map as input cannot be inpainted using this matlab/C++ code.I'm curious about how to carry out this experiment.

[wliusjtu]

When i was reading your paper,i found the experiments conducted using some sturcture missing depth map as input cannot be inpainted using this matlab/C++ code.I'm curious about how to carry out this experiment.

Many thanks for your interest in our paper. For Kinect depth image inpainting, you first need to interpolate these missing depth values as the initial input of our system. Given "IN" as the original depth maps with missing depth values, here is an example of how to interpolate the missing depth values with the "scatteredInterpolant" function in MATLAB:

[x, y] = find(IN>0); % pixel coordinates with valid depth values
[xq, yq] = find(IN==0); %% pixel coordinates with missing depth values
F = scatteredInterpolant(x, y, IN(IN>0));
IN(IN==0) = F(xq, yq);

After these 4 lines of code, you will get an interpolated depth map, then you can use this depth map as the input to our code for depth map refinement.

In addition, we have many papers that conduct depth map restoration, please refer to the following papers whose code can also be found in my GitHub homepage:

1. A generalized framework for edge-preserving and structure-preserving image smoothing, PAMI 2021, AAAI 2020.
2. Semi-Global Weighted Least Squares in Image Filtering. ICCV 2017