Research On Fractional-order Anisotropic Diffusion Models For Image Denoising

Because the noise and edge are both concentrated in the high frequency region of the image,the denoising process often generate artifacts,which will influence the follow-up feature extraction and image understanding. Therefore, it is important to reduce noise while preserving image edges .After summarizing the advantages and disadvantages of the traditional image denoising method, two new denoising models was proposed based on the anisotropic diffusion model. In order to speed up the calculation, the Fourier transform method is used to solve the fractional derivative. In addition, the influence of the differential order on the denoising result and the iterations were explored. The main work is done as follows:(1) Based on the traditional anisotropic diffusion model, a fractional anisotropic diffusion model is proposed by combining the advantages of fractional differentials in edge detection. On this basis, the diffusion function based on the fractional derivative is redesigned. The experimental results show that the method can improve the denoising results and the speed of operation.(2) Based on the fractional anisotropic diffusion model,a mixed diffusion model based on the fractional order diffusion and integer order diffusion was proposed.Further, how to select the parameter was given through the analysis of the experimental results, and the parameter is universal for different images. The denoising results of different images indicated that the proposed model has better denoising performance and fewer iterations than the classical anisotropic diffusion model.Experiments showed that the proposed fractional-order anisotropic diffusion models can improve peak-signal-to-noise ratio at the same time they can fasten the denoising process.