Denoising of Magnetic Resonance and X-ray Images Using Variance Stabilization and Patch Based Algorithms

Developments in Medical imaging systems which are providing the anatomical and physiological details of the patients made the diagnosis simple day by day. But every medical imaging modality suffers from some sort of noise. Noise in medical images will decrease the contrast in the image, due to this effect low contrast lesions may not be detected in the diagnostic phase. So the removal of noise from medical images is very important task. In this paper we are presenting the Denoising techniques developed for removing the poison noise from X-ray images due to low photon count and Rician noise from the MRI (magnetic resonance images). The Poisson and Rician noise are data dependent so they won’t follow the Gaussian distribution most of the times. In our algorithm we are converting the Poisson and Rician noise distribution into Gaussian distribution using variance stabilization technique and then we used the patch based algorithms for denoising the images. The performance of the algorithms was evaluated using various image quality metrics such as PSNR (Peak signal to noise ratio), UQI (Universal Quality Index), SSIM (Structural similarity index) etc. The results proved that the Anscombe transform, Freeman & Tukey transform with block matching 3D algorithm is giving a better result.