A Novel Non-local Image Inpainting Algorithm: Fractal-Based Image Inpainting

Aiming at overcoming the shortage that the searching scope is limited to the source region of the damaged image, a fractal-based digital image in painting algorithm is presented in this paper. Firstly, the source region of the image to be in painted is partitioned into a set of blocks which form a codebook. Secondly, the best-matched block is chosen in the codebook by means of fast fractal theory. In addition, the proportion of confidence term is increased during computing the priority in order to avoid ¿false edge¿ caused by error match and intensify the constraint of searching condition, thus the filling process proceeds by the order of ¿onion-peel¿, meanwhile linear structures can be propagated into the target region. The novel contribution of this paper is to introduce fractal theory into the in painting area and to restore damaged images using the characteristic of self-similarity of the images. Numerical experiments on real and synthetic images are given to illustrate the advantages of the proposed algorithm, and the results compare favorably to those obtained by existing techniques.