A novel image optimization method for dual-energy computed tomography

Dual-energy computed tomography (CT) has been widely used in explosive detection as it has the capability of accurate material discrimination. However, its reconstructions always suffer from severe noise and artifacts. In practical application, the computation of effective atomic number is much more sensitive to the noise and decomposition errors, which significantly degrades its image quality. In this paper we introduced a reference image based non-local image optimization method for dual-energy CT, based on the knowledge that dual-energy reconstructions have strictly the same structure and they are precisely registered. Reconstructed low-energy attenuation image from dual-energy CT serves as the reference image in the proposed method. The structure information was derived from the low-energy attenuation image using a non-local pixel similarity measurement, and weight average the effective atomic number image with the established weight relationship in the corresponding position. Experiments with a commercial dual-layer detector CT scanner demonstrated that the proposed strategy achieves much better results in noise and artifacts reduction than previous methods, and significantly improves the image quality. uction pose a reference image based non-local image optimization method for effective number image in dual-energy CT, which can significantly improve the visual quality and reduce the artifacts.