Denoising low dose CT images via 3D total variation using CUDA

The purpose of this paper is to improve the quality of low dose Computed Tomography (CT) images. Low dose artifacts are the Gaussian noises superimposed on the CT images and are often caused by insufficient calibrated detector and photon starvation. A noise reduction method via three dimensional total variation using Compute Unified Device Architecture (CUDA) is proposed on the three dimensional image. This method can also be employed for low dose noise removal of the two dimensional CT images by decreasing a total variation dimension processing technique. The performance of the proposed algorithm has been tested by using quantitative measures. For quantitative analysis, the quality assessment parameter Peak-signal-to-noise-ratio (PSNR) is used in this paper. Both simulation and real experiment results show that the proposed technique increases the image quality and has high calculation speed.