Comparative study of two image space noise reduction methods for computed tomography: bilateral filter and nonlocal means.

Juan C Ramirez Giraldo, Zachary S. Kelm, Luis S. Guimaraes, Lifeng Yu, Joel Garland Fletcher, Bradley J Erickson, Cynthia H McCollough

Research output: Contribution to journalArticle

Abstract

Optimal noise control is important for improving image quality and reducing radiation dose in computed tomography. Here we investigated two image space based nonlinear filters for noise reduction: the bilateral filter (BF) and the nonlocal means (NLM) algorithm. Images from both methods were compared against those from a commercially available weighted filtered backprojection (WFBP) method. A standard phantom for quality assurance testing was used to quantitatively compare noise and spatial resolution, as well as low contrast detectability (LCD). Additionally, an image dataset from a patient's abdominal CT exam was used to assess the effectiveness of the filters on full dose and simulated half dose acquisitions. We found that both the BF and NLM methods improve the tradeoff between noise and high contrast spatial resolution with no significant difference in LCD. Results from the patient dataset demonstrated the potential of dose reduction with the denoising methods. Care must be taken when choosing the NLM parameters in order to minimize the generation of artifacts that could possibly compromise diagnostic value.

    Fingerprint

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Signal Processing
  • Biomedical Engineering
  • Health Informatics

Cite this