8 February 2017 Recursive Gauss-Seidel median filter for CT lung image denoising
Author Affiliations +
Proceedings Volume 10225, Eighth International Conference on Graphic and Image Processing (ICGIP 2016); 102252D (2017) https://doi.org/10.1117/12.2266968
Event: Eighth International Conference on Graphic and Image Processing, 2016, Tokyo, Japan
Abstract
Poisson and Gaussian noises have been known to affect Computed Tomography (CT) image quality during reconstruction. Standard median (SM) Filter has been widely used to reduce the unwanted impulsive noises. However, it cannot perform satisfactorily once the noise density is high. Recursive median (RM) filter has also been proposed to optimize the denoising. On the other hand, the image quality is degraded. In this paper, we propose a hybrid recursive median (RGSM) filtering technique by using Gauss-Seidel Relaxation to enhance denoising and preserve image quality in RM filter. First, the SM filtering was performed, followed by Gauss-Seidel, and combined to generate secondary approximation solution. This scheme was iteratively done by applying the secondary approximation solution to the successive iterations. Progressive noise reduction was accomplished in every iterative stage. The last stage generated the final solution. Experiments on CT lung images show that the proposed technique has higher noise reduction improvements compared to the conventional RM filtering. The results have also confirmed better anatomical quality preservation. The proposed technique may improve lung nodules segmentation and characterization performance.
© (2017) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Dyah Ekashanti Octorina Dewi, Ahmad Athif Mohd. Faudzi, Tati Latifah Mengko, Koichi Suzumori, "Recursive Gauss-Seidel median filter for CT lung image denoising", Proc. SPIE 10225, Eighth International Conference on Graphic and Image Processing (ICGIP 2016), 102252D (8 February 2017); doi: 10.1117/12.2266968; https://doi.org/10.1117/12.2266968
PROCEEDINGS
5 PAGES


SHARE
Back to Top