Study Of 3D Image Restoration Of The Wide-Field Fluorescence Microscopy

Fluorescence microscopy is widely used in the biological sciences to observe three-dimensional(3D) biological specimens. A standard wide-field fluorescence microscope yields images that are blurred in the axial direction by contributions outside the plane of focus, this problem is solved by the Laser Scanning Confocal Microscope(LSCM). However, LSCM is more complex than the wide-field microscope. And it is usually expensive and has photobleaching problems. As a result, the algorithms of removing the out-of-focus blur in wide-field data have been become a current hot research subject.At present, the image restoration approaches of the fluorescence microscope can be classified as either the linear or the nonlinear methods. The linear methods include nearest neighbor, depth variant PSF, inverse filter, etc. Linear method is a fast approach. To certain extent, it can improve the quality of image. The nonlinear methods include parametric blind deconvolution(PBD), iterative blind deconvolution(IBD), maximum likelihood blind deconvolution(MLBD), etc. Nonlinear algorithm needs iteration and costs much more time, but the quality of images obtained by iterative algorithms is better than that by the linear algorithms.This thesis does some research about the linear and nonlinear methods and acquires some achievements. It is composed of three parts. The first part introduces the basic concept and the image formation theory in a fluorescence microscope, analyses the degradation model of the fluorescence microscope, expounds the ill-posed problemand summarizes the idea that the ill-posed problem is solved by using regularization techniques, the point spread function of the fluorescence microscope, and the main measures to evaluate the image restoration algorithms and several typical image restoration methods for the fluorescence microscope. The second part analyses the principle of the nearest neighbor and presents a new method that combines nearest neighbor with wiener inverse filter. The third part expounds the parametric blind deconvolution algorithm and the model of depth variant PSF, proposes a parametric blind deconvolution algorithm based on depth variant PSF. Both the fourth part and the fifth part use the simulated 3D images and the fluorescence microscope images to verify these two methods.