Parallel Implementation of Collaborative Filtering Technique for Denoising of CT Images

摘要

In the paper parallelization of the collaborative filtering technique for image denoising is presented. The filter is compared with several other available methods for image denoising such as Anisotropic diffusion, Wavelet packets, Total Variation denoising, Gaussian blur, Adaptive Wiener filter and Non-Local Means filter. Application of the filter is intended for denoising of the medical CT images as a part of image pre-processing before image segmentation. The paper is evaluating the filter denoising quality and describes effective parallelization of the filtering algorithm. Results of the parallelization are presented in terms of strong and weak scalability together with algorithm speed-up compared to the typical sequential version of the algorithm.