Hi everyone I meet a problem recently, which is about image processing. The application is image reconstruction.
Originally I want to use ADMM to solve the pixel-based problem which involves gradient and Laplacian matrix. But I find that patch-based image processing is more suitable for image reconstruction.
Therefore, I want to modify the original problem into patch-based one. That is, I divide the original problem into several patches. And then implement ADMM for each patch in parallel. Finally convert the resulting blocks into a big picture.
But I dont know whether this is practicable. Is parallel ADMM for each patch better than pixel-based ADMM? Could anyone provide some suggestions? Thanks in advance!