The objective of this project is to develop a commercial software product to improve the visualization and resolution of two-dimensional images obtained from optical microscopes. The initial focus will be on 2D fluorescence microscopy images and the restoration will be achieved using iterative blind deconvolution techniques. Additional information about the microscope is often unavailable, so the deconvolution will assume nothing about how the image was captured. This will make the product suitable for a wide variety of imaging modalities. However, the 2D blind deconvolution problem is then more difficult than that for current 3D microscope data (where microscope parameters are required) because there are no spatial or spectral constraints on the PSF estimate. The prototype system developed has shown the feasibility of the algorithm. The deconvolution can recover frequency components beyond the diffraction limit of the microscope and allows closely spaced features in the specimen to be more easily resolved. Some of the many potential applications include the enhancement of images from fluorescence in-situ hybridization (FISH), cyto-genetics and experiments using green fluorescent protein (GFP). A particular aim is to make the algorithm as automated and fast as possible, which will allow the processing of time-lapse sequences and previously archived imagery.
The potential market for 2D deconvolution is significantly larger than that for current 3D deconvolution software. Imagery from almost every microscope could benefit because no specialized equipment is required and a diffraction limit is always present. The deconvolution will improve visualization of closely spaced biological structures. The product will be fast, easy-to-use and automated. Details about the experimental setup are not required - enabling archived images to be enhanced. The 2D product will complement existing 3D deconvolution and is most suited to thin specimens, imaged as either a single optical section or from a summed z-stack.
