Recovery of depth information from optical microscope images by constrained deconvolution

A rigorous, new, depth-from-focus algorithm is proposed: microscope images at a series of focal planes are jointly deconvolved, subject to the constraint that the recovered specimen is a two-dimensional surface. A quadratic program is used to estimate the point-spread function of the optical system. Then, depth and brightness are estimated at each location on the surface by minimising a sum of squares plus roughness penalty, using a simulated annealing schedule. The algorithm has been implemented in parallel on a Cray-T3D, and the method is illustrated using brightfield optical sections of a diatom.