Simultaneous total variation image inpainting and blind deconvolution

We propose a total variation based model for simultaneous image inpainting and blind deconvolution. We demonstrate that the tasks are inherently coupled together and that solving them individually will lead to poor results. The main advantages of our model are that (i) boundary conditions for deconvolution required near the interface between observed and occluded regions are naturally generated through inpainting; (ii) inpainting results are enhanced through deconvolution (as opposed to inpainting blurry images). As a result, ringing effects due to imposing improper boundary conditions and errors due to imperfection of inpainting blurry images are reduced. Moreover, our model can also be used to generate boundary conditions for regular deconvolution problems that yields better results than previous methods.