Non-Local Filtering of Poisson-Gaussian Noise in Fluorescence Microscopy Images

• Grant number: 24/03220-6
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): May 01, 2024
• Effective date (End): April 30, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Alexandre Luís Magalhães Levada
• Grantee: Nathalia Brasilino Gimenes
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil

Abstract

Image denoising is a topic of great interest in the areas of image processing and computer vision, especially in medical and biological applications. The most common degradation model in image noise reduction is the Gaussian noise, in which the noise is additive and independent of the signal of interest, which simplifies the derivation of numerical methods and algorithms. However, due to the physical process of fluorescence microscopy images formation, which are widely used to observe small-scale structures, such as bacteria, molecules in cells and the composition of cellular tissues, the signal is corrupted by Poisson-Gaussian noise, which is a mixture of two distinct models. In this research project, as a way to overcome this difficulty, we propose to use non-local filtering algorithms in conjunction with variance stabilization transforms (TEV), in order to adjust the Poisson-Gaussian noise statistics for the Non-local Means filter. Computational experiments with real images will be conducted to evaluate the performance of filtering methods with and without TEVs, in order to compare quantitative image quality metrics. It is expected to achieve better Poisson-Gaussian noise filtering results in fluorescence microscopy images compared to the standard NLM filter, which can be considered a scientific and technological advance in several applications.