High Resolution White Light Diffraction Phase Autofocus Halo-free Imaging Technique

Due to its unique advantages such as high stability,low speckle noise and real-time acquisition and imaging,white light diffraction phase microscopy has been widely concerned and applied in biomedicine,material detection and other fields.However,the traditional white light diffraction phase microscopy technology has low spatial resolution due to the low utilization rate of spatial bandwidth product due to the off-axis technology.Due to the use of white light illumination technology,the recovered phase image has halo phenomenon.It is difficult to focus due to the use of microscopic technology.Although various technologies have been proposed to solve the above problems in recent years,how to achieve high resolution,fast and halo-free self-focusing imaging of white light diffraction phase microscopy is still a scientific and key technical problem to be solved.In this paper,the purpose of high resolution white light diffraction phase self-focusing and halo-free imaging technology is to study a white light diffraction phase microscopy technology based on deep learning,on the basis of the analysis and study of white light diffraction phase shift technology,cosine score self-focusing technology and iterative deconvolution halo removal technology,and make full use of the characteristics of deep learning to extract features from the original data layer by layer to obtain high-level expression.The AU-Net network model is proposed.The neural network model trained by high resolution halo-free focusing phase diagram can realize one exposure acquisition and high resolution and halo-free selffocusing imaging.The main research contents are as follows :By analyzing the fast phase reconstruction algorithm based on the white light diffraction phase microscopy system and the two-step phase shift white light diffraction phase microscopy imaging technology,it is obtained that the combination of the cosine fractional autofocusing algorithm and the two-step phase shift white light diffraction phase microscopy imaging technology can achieve high High-resolution imaging,but the high-resolution image still has the problem of halo effect.In view of the problem that no one has eliminated the halo of the high-resolution phase image,the Hilbert algorithm and the iterative deconvolution algorithm are compared,and the iterative deconvolution algorithm that is more suitable for the original phase to eliminate the halo effect is selected.Resolution phase images were used to eliminate halo effects.The results show that the method has obvious effect but low efficiency.In the white light diffraction phase microscopy system,in view of the low efficiency of iterative deconvolution algorithm to eliminate the halo effect,the improved U-Net neural network model is used to achieve efficient and fast removal of the halo effect.The robustness of the network is verified using a variety of sample data.It lays the foundation for the study of deep learning-based halo-free focusing phase reconstruction.In the research of white light diffraction phase microscopy,the high-resolution two-step phase-shifted white light diffraction phase microscopy technique requires two holograms and requires multiple reconstructions of the focusing algorithm to find the best focus position.Iterative deconvolution algorithm has low efficiency in eliminating halo effect.a high-resolution halo-free focused phase imaging technology,namely the deep learning method,is proposed,which can solve the problems of defocusing,halo effect and imaging at one time.The proposed AU-Net network model has been verified by a variety of sample data to verify the effectiveness and robustness of the proposed network model.