Computational and Optical Methods for Improving Resolution and Signal Quality in Fluorescence Microscopy

We present efficient algorithms for image restoration using Good’s roughness penalty. We assumed Gaussian or Poisson statistics for the noise and derived an algorithm for each case. Performance was tested using simulated three-dimensional imaging with a fluorescence confocal laser scanning microscope. Results were compared to those for algorithms that use Gaussian or entropy penalty terms, which we derived previously [J. Opt. Soc. Am. A 14, 1696–1706 (1997)]. The algorithms based on Good’s roughness yielded superior results. An example is given of the restoration of an image of a biological specimen.