[attachment=5306]
SIMULTANEOUS FRACTAL IMAGE DENOISING AND INTERPOLATION


M. Ghazel R. K. Ward R. Abugharbieh

University of British Columbia
ECE Dept.
Vancouver, B.C. V6T 1Z

E. R. Vrscay , G. H. Freeman
University of Waterloo
Applied Math Dept ECE Dept
Waterloo, ON. N2L 3G1

ABSTRACT

In this paper, a simple and effective fractal-based simultaneous image denoising and interpolation scheme is proposed and implemented. The denoising is performed during the fractal encoding process while the interpolation is performed during the decoding process. The fractal-based image denoising involves predicting the fractal code of the original noiseless image from the statistics of the noisy observation. This fractal code can then be used to generate a fractally denoised estimate of the original image. The fractal interpolation can be easily achieved during the decoding process by iterating the predicted fractal code on a suitably sized blank intital image seed. The cycle spinning algorithm can also be incorporated in the proposed fractal joint denoising and resizing scheme in order to reduce some of the artifacts and enhance the visual quality of the fractally denoised and resized estimates.