Image Denoising Using Fractal Hierarchical Classification
Communications in Computer and Information Science
This paper proposes an efficient yet simple fractal-based image denoising technique. Denoising is carried out during fractal coding process. Hierarchical classification is used to increase encoding speed, and avoid a lot of futile mean-square-error (MSE) computations. Quadtree-based image partitioning using dynamic range and domain sizes is used to increase the degree of noise removal. Further denoising is achieved using pyramidal decoding, using non-arbitrary seed image, and additional post processing. Results from experiments show that our proposed scheme improves the structural similarity (SSIM) index of the Lenna image from 44% to 78% for low noise cases, and from 9% to 35% for high noise cases.
Roy, Swalpa Kumar; Bhattacharya, Nilavra; Chanda, Bhabatosh; Chaudhuri, Bidyut B.; and Banerjee, Soumitro, "Image Denoising Using Fractal Hierarchical Classification" (2018). Conference Articles. 145.