Abstract
Nowadays, 2D image quality of biological tissue has increased drastically because of the progress in image acquisition by microscope. Therefore, 3D superresolution from 2D images is the next aim. If achieved, it can make the structure of the entire biological tissue visualized clearly. Moreover, it is beneficial to the subsequent analysis.This research is intended to increase the resolution of 3D microscopy images which is an interdisciplinary project to build up a cheap but high resolution microscope. The method of this work is to recover genuine 3D biological tissue images from stacks of 2D images. Each 2D biological tissue image is made from confocal fluorescence microscope. By this technique and the subsequent image processing, we intend not only to get clear 2D biological tissue images, but also to reconstruct the genuine 3D biological tissue images.This research focuses on an algorithm for processing a stack of 2D images captured by the microscope from different depth of focuses. Since the point spread functions suffer from the spherical aberration in our microscope, they are no more space-invariant models. Therefore, the research is focused on how to modify theoretical point spread functions to adapt to our capturing image system and obtain deblurred images by deconvolution. In the research, the mathematical models of depth-variant point spread functions will be formulated. Furthermore, the depth-variant deconvolution algorithm that incorporates the depth-variant point spread functions will be introduced.