Denoising of low dose CT image with context-based BM3D

摘要

Low-dose CT scanning can efficiently reduce the amount of radiation to the patient, which leads to a decline in image quality and influences the doctor's diagnosis of lesions. An improved block-matching and 3D filtering (BM3D) based on the context is proposed to reduce the noise of low-dose CT and improve the image quality. Further, a visual attention method is used to highlight the lesions so that the diseased tissues can be improved in imaging contrast. The Contrast-to-noise ratios (CNR) of 2 regions-of-interest were improved from 0.8 and 0.7 to 1.2 and 2.5, respectively for two patients. For the liver patient, the difficult-to-see lesions in the original CT image became highly visible in our study.