Confidence Weighted Dictionary Learning algorithm for low-dose CT image processing

Though clinically desirable, Computed Tomography (CT) images tend to be severely degraded by excessive noise and artifacts. This paper proposes a novel post-processing approach termed Confidence Weighted Dictionary Learning (CW-DL) to improve low-dose CT (LDCT) images. The proposed CW-DL algorithm introduces a novel intensity constrained strategy into the frame of dictionary learning (DL) processing, and demonstrates an improved performance in artifact suppression. Experiment results show that the proposed CW-DL algorithm can lead to effective suppression of both mottled noise and artifacts in LDCT images.