Three-dimensional image reconstruction of fluorescence micrographs without knowing the point spread function

Abstract

The term blind deconvolution refers to the deconvolution, or deblurring, of a signal (optical, sound, or other) without explicit knowledge of the point spread function (PSF). The PSF is, instead, reconstructed concurrently with the deblurred signal from the collected, noisy data. Blind deconvolution methodologies have been under study for general signal processing applications, such as in restoring phonographs, as early as 1968. Ayers and Dainty have performed pioneering blind deconvolution research for deblurring 2D images, which has been applied to astronomy. Inspired by this earlier blind deconvolution research, we have taken a new approach which is based on maximum likelihood estimation (MLE). Our approach is extended from previous research in applying MLE to positron-emission tomography (PET). A novelty of our research is in the application of MLE and blind deconvolution to 2D and 3D fluorescence microscopy.The fundamental advantage of the MLE approach over some of the other blind deconvolution approaches is that it is a mathematical optimization approach, wherein the likelihood functional of the collected image data, the fluorescence probe concentration and the PSF is maximized.