Denoising techniques for multi-parametric prostate MRI: A Comparative Study

Since recent years, denoising become one Of the most active area of research in image processing topic. Usually, MR images are affected by noise and artifacts during the acquisition process. Therefore, many denoising algorithms have been developed although noise elimination still an undefended challenge. In this paper, we study firstly different denoising filters for T2-Weighted prostate cancer MR images, in order to select the appropriate filter. As example of denoising filters, homomorphic, Median, Wavelet, nonlocal means, gaussian, Anisotropic, Laplacian, Cure-LET, LMMSE and bilateral. Then, we discuss the problem of evaluation of image quality which become necessary. As example of evaluation metrics, we present the PSNR, MSE and SSIM. We consider both subjective and objective quality assessment parameters for determining a final score of filters executed over 40 T2-Weighted MR images. This study concludes that Anisotropic filter should be opted for denoising T2-Weighted MR image since its details preserving capability.