Fourier Synchrosqueezed Transform When Used for Medical OCT Image Enhancement

A projection scheme in the Fourier domain is proposed to suppress speckle noise in Optical Computed Tomography (OCT) images. A comparison is held between the discrete fast Fourier transform (FFT) and the discrete Fourier synchrosqueezed transform (FSST) through two hard thresholding techniques; the projection of signal space onto the noisy subspace in the frequency domain, and the selection of large magnitude coefficients from the spectrum. Results show the efficacy of the Fourier transform compared to the Fourier synchrosqueezed transform, and that is through the projection method over the selection of large magnitude coefficients method. When applied to simulated and real OCT images, and in terms of numerical and visual assessment, the proposed scheme has outperformed benchmark schemes used in the area of OCT image despeckling, such as Total Variation (TV) filter, Geometric Non-Linear Diffusion (GNLD) filter and Detail Preserving Anisotropic Diffusion (DPAD) filter.