Projection domain denoising method based on dictionary learning for low-dose CT image reconstruction

Subtitle:

Article type: Research Article

Authors: Zhang, Haiyan^a | Zhang, Liyi^{a; b} | Sun, Yunshan^{a; b; *} | Zhang, Jingyu^a

Affiliations: [a] School of Electronic Information Engineering, Tianjin University, Tianjin, China | [b] School of Information Engineering, Tianjin University of Commerce, Tianjin, China

Correspondence: [*] Corresponding author: Yunshan Sun, School of Information Engineering, Tianjin University of Commerce, Jinba Road, Beichen District, Tianjin 300134, China. Tel.: +86 022 26675771; E-mail:[email protected]

Abstract: Reducing X-ray tube current is one of the widely used methods for decreasing the radiation dose. Unfortunately, the signal-to-noise ratio (SNR) of the projection data degrades simultaneously. To improve the quality of reconstructed images, a dictionary learning based penalized weighted least-squares (PWLS) approach is proposed for sinogram denoising. The weighted least-squares considers the statistical characteristic of noise and the penalty models the sparsity of sinogram based on dictionary learning. Then reconstruct CT image using filtered back projection (FBP) algorithm from the denoised sinogram. The proposed method is particularly suitable for the projection data with low SNR. Experimental results show that the proposed method can get high-quality CT images when the signal to noise ratio of projection data declines sharply.

Keywords: CT image reconstruction, dictionary learning, projection data denoising, low-dose CT

DOI: 10.3233/XST-150509

Journal: Journal of X-Ray Science and Technology, vol. 23, no. 5, pp. 567-578, 2015