Inpainting using geometrical grouplets

In this paper, we propose a two-steps image inpainting algorithm that relies on geometrical grouplets. The grouplet orthogonal bases, that were introduced by Mallat in [1], are constructed with multiscale association field that groups pixels to take advantage of geometrical image regularities. These bases are used to represent the complex geometry of the image to be inpainted. Having well represented the image geometry by using geometrical grouplets, missing data are synthesized by propagating the geometrical information from outside to the inside of the inpainting zone. The first step of the algorithm consists of computing the multiscale association field and the grouplet transform of the image. In the second step, the integral curves of the association field are computed and propagated inside the inpainting zone to fill in the missing data. This step is accomplished by minimizing a functional whose role is to reconstruct the missing or damaged zone independently of the size and topology of the inpainting domain. Extensive experiments are carried out in order to validate and compare the algorithm both quantitatively and qualitatively. They show the advantages of our algorithm and its readily application to real world cases.