questions for the code of 2D inpainting "reconstruct_2D_subsampling.m"

[codemonkey123-coder]

reconstruct_2D_subsampling.m

% % %Sampling matrix
MtM = ones(size(signal));
% % %MtM(1:2:end, 1:2:end) = 1;
MtM(rand(size(MtM)) < 0.5 ) = 1;
signal_sparse = signal;
signal_sparse( ~MtM ) = 0;

---

From the code, we can find that the matrix MtM is initialized as ones. MtM(rand(size(MtM)) < 0.5 ) = 1; means that the missing part in the mask is ones. Above all, the mask doesn't work, so the signal_sparse is same as signal. Then we can find that the inpainting process is same as the process of Convolutional Sparse Coding, it doesn't do anything with inpainting. Then, i initalize the MtM as zeros. The inpainting results of the code are bad, it is quite different from the results in the paper. I want to know how we can get the great inpainting results in the paper.