New variational image decomposition model for simultaneously denoising and segmenting optical coherence tomography images

Research output: Contribution to journalArticlepeer-review


  • Jinming Duan
  • Christopher Tench
  • Irene Gottlob
  • Frank Proudlock
  • Li Bai

Colleges, School and Institutes

External organisations

  • University of Nottingham
  • University of Leicester


Optical coherence tomography (OCT) imaging plays an important role in clinical diagnosis and monitoring of diseases of the human retina. Automated analysis of optical coherence tomography images is a challenging task as the images are inherently noisy. In this paper, a novel variational image decomposition model is proposed to decompose an OCT image into three components: the first component is the original image but with the noise completely removed; the second contains the set of edges representing the retinal layer boundaries present in the image; and the third is an image of noise, or in image decomposition terms, the texture, or oscillatory patterns of the original image. In addition, a fast Fourier transform based split Bregman algorithm is developed to improve computational efficiency of solving the proposed model. Extensive experiments are conducted on both synthesised and real OCT images to demonstrate that the proposed model outperforms the state-of-the-art speckle noise reduction methods and leads to accurate retinal layer segmentation.


Original languageEnglish
Pages (from-to)8901-8922
Number of pages22
JournalPhysics in Medicine and Biology
Issue number22
Publication statusPublished - 4 Nov 2015


  • fast Fourier transform, image decomposition, image segmentation, noise removal, Optical coherence tomography, split Bregman algorithm, variational methods