Denoising of biomedical images using two-dimensional fourier-bessel series expansion-based empirical wavelet transform

Abstract

Image denoising is a kind of image enhancement, which is an initial step for biomedical image processing and analysis. The motivation for denoising is that an enhanced image provides more accurate analysis, interpretation, and classification. In this chapter, order-zero and order-one two-dimensional Fourier-Bessel series expansion-based empirical wavelet transform (2D-FBSE-EWT) methods are used with a wavelet thresholding approach to denoise images. Order-zero and order-one 2D-FBSE-EWT methods are variants of 2D-EWT that use the order-zero and order-one Fourier-Bessel series expansion (FBSE) spectrum instead of the Fourier transform (FT) spectrum. In a wavelet thresholding approach, adaptive thresholds are estimated for each sub-image coefficient, followed by hard thresholding of the coefficients. After thresholding of coefficients, all modified sub-images are added to get a denoised image. Peak signal-to-noise ratio is used as a performance measure parameter. For analysis purposes, three biomedical images, fundus, mammogram, and skin images, with additive white Gaussian noise of standard deviations 10, 20, 30, and 40 are used. The results of order-zero and order-one 2D-FBSE-EWT methods are then compared with the result of the 2D-empirical wavelet transform (2D-EWT). The order-zero 2D-FBSE-EWT provided better performance as compared to order-one 2D-FBSE-EWT and 2D-EWT methods. © 2022 Taylor & Francis Group, LLC. All rights reserved.