Compensation of the Light Attenuation with Depth of Images Captured by a Confocal Microscope Using a MRF Deformation Model and Graph Cuts

Abstract

We study series of fluorescent optical sections, i.e. three-dimensional (3D) biomedical images, captured by a confocal laser scanning microscope (CLSM). Fluorescent image intensities of optical sections from deep layers of the specimen are often darker than intensities of sections from the top layers due to absorption and scattering of both excitation and fluorescent light. To solve this problem we apply a Markov Random Field (MRF) model including an efficient deformation model for tracking structures within the 3D images for computation of optical flow. We approach the corresponding optimization problem by the graph cuts. Image intensities of optical sections are recomputed according to the found optical flow, since the flow gives us evaluation of their proper brightness. Finally, the light attenuation with depth is compensated by matching accumulative histograms of optical sections of the original series with respect to optical sections improved by the optical flow. By this approach we obtain an algorithm that is less sensitive to changes of structures within series (especially to their enlargement and diminishing) than algorithms based purely on histogram matching, warping or equalization.