Directional bilateral filters for smoothing fluorescence microscopy images

Abstract

Images obtained through fluorescence microscopy at low numerical aperture (NA) are noisy and have poor resolution. Images of specimens such as F-actin filaments obtained using confocal or widefield fluorescence microscopes contain directional information and it is important that an image smoothing or filtering technique preserve the directionality. F-actin filaments are widely studied in pathology because the abnormalities in actin dynamics play a key role in diagnosis of cancer, cardiac diseases, vascular diseases, myofibrillar myopathies, neurological disorders, etc. We develop the directional bilateral filter as a means of filtering out the noise in the image without significantly altering the directionality of the F-actin filaments. The bilateral filter is anisotropic to start with, but we add an additional degree of anisotropy by employing an oriented domain kernel for smoothing. The orientation is locally adapted using a structure tensor and the parameters of the bilateral filter are optimized for within the framework of statistical risk minimization. We show that the directional bilateral filter has better denoising performance than the traditional Gaussian bilateral filter and other denoising techniques such as SURE-LET, non-local means, and guided image filtering at various noise levels in terms of peak signal-to-noise ratio (PSNR). We also show quantitative improvements in low NA images of F-actin filaments.

ACDC: Automated Cell Detection and Counting for Time-lapse Fluorescence Microscopy

“…This simple yet effective strategy allows for contrast enhancement [44]. Bilateral filter has been shown to work properly in fluorescence imaging even preserving the directional information, such as in the case of the F-actin filaments [45]. This denoising technique was effectively applied to biological electron microscopy [46], as well as to cell detection [47], revealing better performance—compared to low-pass filtering—in noise reduction without removing the structural features conveyed by strong edges.…”

MedGA: A novel evolutionary method for image enhancement in medical imaging systems

“…This simple yet effective strategy allows for contrast enhancement [548]. Bilateral filter has been shown to work properly in fluorescence imaging even preserving the directional information, such as in the case of the F-actin filaments [631]. This denoising technique was effectively applied to biological electron microscopy [295] as well as to cell detection [359], revealing better performance, with respect to low-pass filtering, in noise reduction without removing the structural features conveyed by strong edges.…”

Preface to Special Topic: Emerging Techniques in Fluorescence Microscopy and Imaging

“…Specifically the papers published in this special issue include, fast fluorescence imaging techniques (Jabbar et al, 21 Cha et al 22 ), Super-resolution (You et al 23 ), portable Microscopy (Jagannadh et al 24 ), Multiphoton Microscopy (Ehmke et al 25 ), Image Reconstruction Techniques (Venkatesh et al 26 ) and application in the field of Medicine/Biology (Mandal et al, 27 Rai et. Al.…”

“…The second technique by Jabbar et al 21 employ dedicated multi-core computational engines for 3D imaging with a speed approximately 200 times faster than existing central processing unit (CPU) based computing engines. Ehmke et al 25 report an optimized and efficient multiphoton microscopy technique for imaging highly-ordered biological tissue whereas, Venkatesh et al 26 propose signal processing techniques such as, Directional filters for high quality (noise-free and edge-preserving) fluorescence imaging. Jagannadh et al 24 proposed and demonstrated low-cost and portable variant of optical microscope.…”