The impact of improved non-local means denoising algorithm on photon-counting X-ray images using various Al additive filtrations

Abstract

The filters used in X-ray systems affect the image quality and radiation dose. Although using filters has advantages in terms of image quality and dose, image noise increases because the amount of X-rays in the energy range of the entire area is reduced. Thus, a newly improved non-local means (INLM) denoising algorithm was modeled and applied to X-ray images based on the thickness of the aluminum (Al) filter, which is most commonly used in X-ray imaging systems, to prove its potential use. For X-ray image acquisition, a high-performance cadmium telluride material-based detector and a tube containing Al filters (1-, 3-, and 5-mm thickness) were designed. The proposed INLM denoising algorithm was modeled by including the improved weights for the gradient information for each pixel in the conventional NLM-based equation. When the proposed INLM denoising algorithm was applied to the acquired photon-counting X-ray images, results showed superior performance for both noise characteristics and no-reference-based image evaluation. In particular, we confirmed that when the INLM was applied to X-ray images using 5 mm Al thickness, noise characteristics and no-reference-based evaluation results were improved by approximately 1.36 times and 1.06 times, respectively, compared to the conventional NLM. The study proved that choosing the INLM in photon-counting X-ray images using 5 mm-thick Al filtration is vital for the success of image processing applications. © 2021 Elsevier B.V.