Three-dimensional (3D) fluorescence microscopy is a powerful tool for the study of living specimens. 3D images are normally obscured by out-of-focus blur and often artifactual because of the limited band of spatial frequencies that the conventional fluorescence microscope can image. These problems can be ameliorated by optical and/or computational methods. The development of newer microscopes for 3D imaging has been fast paced resulting on a wide variety of approaches to overcome the out-of focus blur and the band limitation of conventional microscopes. These approaches include confocal and partially confocal scanning microscopes, tow-photon and three- photon fluorescence excitation, standing wave, among other. These approaches improve the imaging properties but are not blur-free and computational approaches may further improve image quality. Despite the widespread availability of computers attached to microscopes, computational deconvolution methods have not enjoyed the fast and rich development of the optical instruments. There are vary few commercially available deconvolution packages and even those are based on the first few algorithms develop for 3D deconvolution. These approaches improve the imaging properties but are not blur-free and computational approaches may further improve image quality. Despite the widespread availability of computers attached to microscopes, computational deconvolution methods have not enjoyed the fast and rich development of the optical instruments. There are very few commercially available deconvolution packages and even those are based on the first few algorithms developed for 3D deconvolution. These algorithms are only partially successful and significant improvements can be obtained with algorithms that are based on a thorough mathematical model for the microscope and the light detector. Furthermore, a systematic evaluation of the exiting algorithms has not been done and thus it is unclear what it is possible to achieve with them. The long-term goal of the research proposed is to further develop computational deconvolution algorithms that account for the most significant sources of degradation in the image and to provide guidelines for the use of the algorithms and for their capabilities and limitations. To achieve this goal we will first derive algorithms based on an accurate model for the microscope and light detector then we will do a thorough evaluation of these and other algorithms to assess their relative merits and to provide guidelines for their use.
