Alternating direction method of multipliers applied to 3D light sheet fluorescence microscopy image deblurring using GPU hardware

Abstract

This paper focuses on the deblurring and denoising of Poisson noise contaminated images acquired with a new imaging technique producing large 3D data sets: Light Sheet Fluorescence Microscopy. This paper details the optimization algorithm used, which is based on the Alternating Direction Method of Multipliers, and its efficient implementation using GPU hardware. In practice, a 3D 100 million voxel image is deconvolved in five minutes, which is at least 25 times faster than a state-of-the-art MATLAB implementation.