A New Spectral Intrinsic Decomposition based Image Inpainting Method

In this paper, an inpainting approach derived from a multiscale and adaptive analysis is proposed. The proposed method use some pixels as characteristics points of the Spectral Intrinsic Decomposition (SID) of inputed image and compute missing pixels by the recomposed image of components witch eigenvalues are equal to one. The recomposition is driven by a diffusivity function. The inpainting method is implemented and applied to well known images data base. The results shows the effectiveness of the new method by reducing considerably noise and gives promising results for image restoration.