The mathematical interest in this field became increasingly active in the last decade since the very first works on image interpolation by Mumford, Nitzberg and Shiota [22], Masnou and Morel [21] and Caselles, Morel and Sbert [5]. However it was the pioneering work of Bertalmio et al [2] who proposed an algorithm to imitate the work of inpainting artists who manually restore old damaged pictures which mainly motivated all the subsequent research in this field [11, 12]. This algorithm cleverly transports a smoothness image measure (namely the Laplacian of the image) along the level lines (contours of the same image intensity) directed into the inpainting domain; in their paper, they also showed that some intermediate steps of anisotropic diffusion are necessary to avoid blurring of edges. This algorithm was created mostly intuitively but later on turned out to be closely related to the Navier-Stokes equation, as showed by Bertozzi et al; see [4]. Since then, many other authors have proposed different models for digital inpainting.