Complex-domain SVD- and sparsity-based denoising for optical diffraction tomography

• Published in: Optics and lasers in engineering 2022-12, Vol.159, p.107228, Article 107228
• Main Authors: Shevkunov, Igor, Ziemczonok, Michał, Kujawińska, Małgorzata, Egiazarian, Karen
• Format: Article
• Language: English
• Subjects: Noise filtering; Noise in imaging systems; Optical diffraction tomography; Refractive index reconstruction; Singular value decomposition; Sparse representation; Tomographic imaging
• ISSN: 0143-8166; 1873-0302
• DOI: 10.1016/j.optlaseng.2022.107228

•Complex-domain sparse representation and SVD provide high quality denoising in optical diffraction tomography.•Proposed approach outperforms traditional slice-by-slice filtering techniques, 3D median filter, and BM4D.•Proposed denoising algorithm enables imaging in low signal-to-noise regime. [Display omitted] In this paper, we adopt a complex-domain cube filter (CCF) developed for hyperspectral 3D complex domain images for noise suppression of 3D complex-valued data in optical diffraction tomography. CCF is based on two processing steps: singular value decomposition (SVD) and complex-domain sparsity-based filter (CDID). SVD provides data compression and CDID noise suppression in the compressed domain. We demonstrate that the CCF algorithm can be used to denoise captured projections (sinogram), which results in enhanced tomographic reconstruction. The accuracy and quantitative advantage of CCF application are shown in simulation tests and in the processing of the experimental data. We show that the algorithm effectively suppresses noise and retrieves objects’ details even for highly noisy data.