Double Noise Filtering in CT: Pre- and Post-Reconstruction

Motivated by the ALARA (As Low As Reasonably Achievable) principle, this paper proposes to denoise Computed Tomography (CT) images by using a double-filtering approach. First, projection data were filtered using methods to filter Poisson noise (pre-filtering step). Then the filtered back projection (FBP) algorithm was applied to image reconstruction. After, the reconstructed images were denoised by using suitable methods for filtering Gaussian noise (post-filtering step). Finally, known metrics of image quality evaluation (such as SSIM and PSNR) were used to compare the filtered images with the ones considered ideal images in various combinations of filters. The results lead to the conclusion that a second filtering applied on image domain can improve the CT denoising quality from pre-filtering step. Thus, CT double-filtering strategy achieved a better balance between noise reduction and details preservation.