Multi-scale fractal coding for single image super-resolution

Abstract

Fractal image coding can specially utilize spatial information and self-similarity structural information of an image to achieve image super-resolution. However, fractal image coding based on single scale brings the problems of block effect and loss of details. In this paper we propose a multi-scale fractal coding method for single image super-resolution. The proposed method integrates the fractal results of different scales and uses back-projection to further optimize the result. Experimental results show that the proposed method can remove the block effect, improve the loss of details and keep smooth of flat area and sharpness of edges in the reconstructed image. Compared with conventional fractal coding and cubic splines interpolation, our method is superior to both of them subjectively and objectively. © 2014 Springer International Publishing Switzerland.