Total Variation Denoising for Optical Coherence Tomography

Michael Shamouilian, Ivan Selesnick

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper introduces a new method of combining total variation denoising (TVD) and median filtering to reduce noise in optical coherence tomography (OCT) image volumes. Both noise from image acquisition and digital processing severely degrade the quality of the OCT volumes. The OCT volume consists of the anatomical structures of interest and speckle noise. For denoising purposes we model speckle noise as a combination of additive white Gaussian noise (AWGN) and sparse salt and pepper noise. The proposed method recovers the anatomical structures of interest by using a Median filter to remove the sparse salt and pepper noise and by using TVD to remove the AWGN while preserving the edges in the image. The proposed method reduces noise without much loss in structural detail. When compared to other leading methods, our method produces similar results significantly faster.

Original languageEnglish (US)
Title of host publication2019 IEEE Signal Processing in Medicine and Biology Symposium, SPMB 2019 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728143439
DOIs
StatePublished - Dec 2019
Event2019 IEEE Signal Processing in Medicine and Biology Symposium, SPMB 2019 - Philadelphia, United States
Duration: Dec 7 2019 → …

Publication series

Name2019 IEEE Signal Processing in Medicine and Biology Symposium, SPMB 2019 - Proceedings

Conference

Conference2019 IEEE Signal Processing in Medicine and Biology Symposium, SPMB 2019
CountryUnited States
CityPhiladelphia
Period12/7/19 → …

Keywords

  • Median Filtering
  • Optical Coherence Tomography
  • Total Variation Denoising

ASJC Scopus subject areas

  • Signal Processing
  • Health Informatics

Fingerprint Dive into the research topics of 'Total Variation Denoising for Optical Coherence Tomography'. Together they form a unique fingerprint.

Cite this