Reweighted l2 norm minimization approach to image inpainting based on rank minimization

This paper proposes a rank minimization based approach to a novel image painting. We utilize the 2-D autoregressive (AR) model to describe the image data, and formulate the image inpainting problem as the system identification problem of finding the minimum order system. This problem is described as the rank minimization problem, which is NP hard in general. To solve the problem approximately, this paper proposes a fast algorithm based on the iterative reweighted least square (IRLS). Numerical examples show that the proposed algorithm recovers missing pixels well.