Convolutional dictionary learning for blind deconvolution of optical coherence tomography images

• DOI: 10.1364/boe.447394
• Publication Date: 2022-02-10T07:30:10Z
• Authors: Junzhe Wang; Brendt Wohlberg; R. B. A. Adamson
• Keywords: Point spread function; Speckle noise; Convolution (computer science)

In this study, we demonstrate a sparsity-regularized, complex, blind deconvolution method for removing sidelobe artefacts and stochastic noise from optical coherence tomography (OCT) images. Our estimates the complex scattering amplitude of tissue on line-by-line basis by estimating deconvolving one-dimensional axial point spread function (PSF) measured OCT A-line data. We also present strategy employing sparsity weighting mask to mitigate loss speckle brightness within tissue-containing regions caused sparse deconvolution. Qualitative quantitative analyses show that approach suppresses background better than traditional spectral reshaping techniques, with negligible structure. The technique is particularly useful emerging applications where images contain strong specular reflections at air-tissue boundaries create large artefacts.