A novel diffusivity function-based image denoising for MRI medical images

Medical imaging is essential for accurate diagnosis. In medical imaging, various algorithms for image denoising have been developed. However, some drawbacks have been identified, including the blocking effect, which results in excessive smoothing of the images, and the loss of image detail. To generate noise on images, this article used Poisson noise. We propose a new diffusivity function-based partial differential equation method used for image denoising with the aid of exploiting the statistical properties of observed noisy images. This model involves a Quaternion Wavelet Transform, which is responsible for creating the different coefficients of a noisy image. Utilizing the soft threshold function, an improved generalized cross-validation function is responsible for determining the best threshold value. This optimal threshold value is then used to control the diffusion process by means of a new diffusivity function. Here, we introduce the fourth-order partial differential equation diffusivity function, an unique diffusion coefficient that is more effective than earlier approaches at eliminating noise and maintaining edges. Finally, the experiments of the proposed method are measured using the peak signal-to-noise ratio (42.78 dBs), mean square error (3.4206), structural similarity index (99.645 %), and standard error (peak signal-to-noise ratio (40.677 dBs), mean square error (5.867), and structural similarity index (97.978 %)) as well as compared to the results of other conventional image denoising techniques (improved partial differential equation-based total variation model, Generalization cross-validation with diffusivity function, non-linear nonlocal diffusion equation, Efficient anisotropic diffusion model, and Hessian matrix-based fourth-order anisotropic diffusion filter). The proposed method produces superior qualitative and quantitative results. The MATLAB R2020a version was used to analyze the results.