Performance comparison of wavelet families for noise reduction and intensity thresholding in Fourier Ptychographic microscopy

Microscopy is going through a digital renaissance and new schemes are developed where computer and algorithms constitute an integral part of the imaging process itself. Computational microscopy increases performance by offering better resolution, larger field of view, quantitative contrast and also reduced size, weight and economic cost. Fourier Ptychographic microscopy utilizes multiple images of a sample taken at lower resolution, each illuminated with a unique incidence angle coherent source, and synthesizes one high resolution complex valued image by iterative phase retrieval algorithms. The recorded images are often corrupted with background noise and pre-processing is needed to improve the quality of the FP recovered image. The pre-processing involves data denoising, thresholding and intensity balancing. We have investigated different wavelet families to test their performance in terms of having compact support and giving the desired level of decomposition for optimal intensity thresholding and denoising in Fourier Ptychography (FP). The wavelet families Daubechies, Biorthogonal, Reverse Biorthogonal, Coiflet, Fejer-Korovkin, Discrete Meyer and Symlet with different compact support have been studied. The obtained threshold was used with noisy synthetic and experimental images for a variety of objects to evaluate the performance of the described framework. In particular, Reverse Biorthogonal wavelets were found to preserve useful signal in corrupted images to a great extent (RMS error 0.39) with low computational cost. Consequently, quantitatively more correct amplitude and phase images with uniform and homogeneous background could be recovered. • In FPM, Experimental images contain background noise and pre-processing is always needed. • Pre-processing involves data denoising, thresholding and intensity balancing. • Different wavelet families have been investigated to evaluate performance for denoising in FPM • Daubechies, Biorthogonal, Reverse Biorthogonal, Coiflet, FejerKorovkin, Discrete Meyer and Symlet wavelets have been studied. • Reverse Biorthogonal wavelets were found to preserve useful signal in corrupted images to a great extent.