A software-based method for eliminating grid artifacts of a crisscrossed grid by mixed-norm and group-sparsity regularization in digital radiography

Abstract

X-ray grids are typically used in clinical practice to improve image quality by decreasing the effects of scatter radiation reaching the image detector. Recently, we developed a prototype crisscrossed grid with a strip density of 3.448 lines/mm (nominal) by adopting a micro-controlled sawing process and carbon interspace material to further improve its scatter radiation removal ability. However, the most critical obstacle remaining for the successful use of X-ray grids in digital radiography is the observation of grid artifacts, such as grid shadows and moiré, which can result in a misdiagnosis by physicians. Accordingly, in this study we propose a new software-based method using an alternative minimization iteration via mixed-norm and group-sparsity regularization to efficiently eliminate grid artifacts of a crisscrossed grid in digital radiography. We performed a numerical simulation with synthetic images degraded by grid shadows of 0.2, 0.9, and 1.4 lines/mm and an experiment with two real X-ray images degraded by a moiré of 1.196 lines/mm to verify the efficacy of the proposed method. We compared the image restoration performance of the proposed method to those of other existing approaches to demonstrate its competitive efficacy. Our results indicate that the proposed method can preserve grid-artifact-free object information in the image restoration process.