Light field high resolution deconvolution method and system based on convolutional neural network

Abstract

The invention discloses a method and system for high-resolution deconvolution of a light field based on a convolutional neural network, wherein the method includes: collecting light field microscopic data through a light field microscopic imaging system; According to the distribution characteristics and parameters of the distribution data set, the distribution data set of the simulated sample is generated; the simulation imaging of the experimental light field system is carried out on the above data set according to the principle of light field imaging, and the simulation light field image is obtained by introducing experimental factors; Based on the Lucy deconvolution algorithm, the three-dimensional deconvolution algorithm of the light field is realized, and the simulated point spread function is used as the prior information to deconvolute the simulated light field image to obtain the preliminary reconstruction volume distribution data; generate a deep convolutional neural network ; Input the results of the preliminary reconstruction into the deep convolutional neural network to obtain the predicted value of the corresponding sample volume distribution data. This method overcomes some inherent problems in traditional light field 3D reconstruction algorithms, and realizes 3D reconstruction of light field data with high resolution and less artifacts.

Description

technical field [0001] The invention relates to the technical fields of computational optics, computational photography, computer vision and computer graphics, in particular to a method and system for high-resolution deconvolution of light fields based on convolutional neural networks. Background technique [0002] The three-dimensional rapid dynamic imaging of living cells and tissues is an important issue concerned by modern life science and medical technology research institutes, and there is an urgent need for high-resolution rapid three-dimensional imaging technology in related research. Light field microscopic imaging technology provides a feasible solution to the demand for fast imaging by virtue of its ability to simultaneously collect spatial information and angular information. After being introduced into the field of optical microscopy, light field microscopy has played an important role in many biological imaging problems such as calcium signal imaging. [0003]...

AI Technical Summary

Problems solved by technology

[0003] Although light field microscopy has achieved a series of excellent results, there are still some difficult problems that limit the further expansion of its application range.

In order to improve the imaging speed, the light field microscopy imaging technology sacrifices the spatial resolution for the angular resolution. Although the deconvolution algorithm can compensate the sacrificed spatial resolution to a certain extent, it is still far lower than that of the optical system. Diffraction limit, and the deconvolution reconstruction algorithm used will introduce new problems while improving the spatial resolution

First, the deconvolution algorithm requires a more accurate estimation of the point spread function of the realization system, which is difficult to measure directly in the experiment; second, the three-dimensional deconvolution algorithm requires a large number of iterative steps to achieve a better The convergence effect has caused a huge computational cost; third, due to the sampling rate limitation at the focal plane, the resolution near the focal plane cannot be effectively improved by the deconvolution algorithm; fourth, due to the ill-conditioned nature of the problem, the deconvolution The product algorithm often introduces unpredictable noise, which has a great impact on the reconstruction results

Images