Microscope image deblurring method based on dense network

Abstract

The invention discloses a microscope image deblurring method based on a dense network. Through adoption of a deep learning method, a deblurring relationship between a blurred acquired image of the fluorescence microscope and sample estimation obtained by adopting a Richardson-Open deconvolution algorithm is learned, so that deblurring processing of data acquired by the fluorescence microscope is realized. According to the invention, a label of a network structure is obtained by using a traditional Richardson-Shanxi deconvolution algorithm, so that the reliability of network learning is ensured; when the test data is input into the built general dense network, the output can better deblurring the fluorescence microscope, the efficiency of deblurring by using the dense network is higher thanthe efficiency of deblurring by using the traditional joint deconvolution algorithm, and a non-iterative deblurring mode is adopted, so that the problem that the iteration stop standard of the traditional iterative algorithm is difficult to select is solved.

Description

technical field [0001] The invention belongs to the technical field of fluorescence microscopic imaging, and in particular relates to a method for deblurring a microscope image based on a dense network. Background technique [0002] Fluorescence microscopy is an important imaging technology in the biological field. Its application range extends from single-cell imaging to large-scale tissue imaging; by staining biological samples with fluorescent proteins, and using lasers to excite biological samples with fluorescence, a submicron space can be obtained. Resolution, molecular specificity, and high-contrast biological images, these properties enable fluorescence microscopy to be applied to the study of intracellular material uptake, transport, distribution, and localization, enabling researchers to directly explore the structure and function of biological samples. However, during the imaging process, due to the limitations of the diffraction limit of various lenses, as well a...

AI Technical Summary

Problems solved by technology

None of these methods deviated from the Richardson-Lucy deconvolution framework. The main problems are: 1. The selection of the point spread function will affect the effect of the algorithm; 2. The selection of the number of iterations cannot be solved; Therefore, how to obtain a new algorithm that does not consider the point spread function and the number of iterations is a hot topic in this field.

Images