Volumes acquired through conventional optical microscopes have two main sources of degradation: a) frequencies cut-off that works as a low-pass filter, which causes a blurring effect especially throughout the z-axis (Goodman, 1996), and b) photon-count noise, a signal-dependent noise that can be well modeled by a Poisson distribution, due to the nature of light based sensors, such as the CCD (charged-coupled devices) (Snyder & Miller, 1991).
Pre-processing steps in computer vision applications uses often linear and smoothing filters to improve image condition and the posterior analysis (Colicchio et al., 2005). However, these linear filters and smoothing operators such as Gaussian filters, as well as non-linear morphological opening or closing operators, may remove important structures present in images (Agard, 1984). Then, restoration methods that are based on image degradation and formation can improve the results. By using the theoretical model of a point-spreading function (PSF) of the microscope developed by Gibson and Lanni (1991), and using the noise statistics knowledge, it is possible to apply non-linear restoration algorithms that are best suited to the problem.