Research On Fourier Ptychographic Microscopic Reconstruction Method Based On Feature Fusion Of Multi-Convolution Network

Fourier ptychographic microscopy imaging technology,is the use of Fourier ptychographic microscopy imaging system to acquire low-resolution intensity images.Combined with reconstruction algorithms to reconstruct high-resolution magnitude images and phase images from the acquired low-resolution images,it solves the contradiction of not being able to balance large field of view with high resolution in microscopic imaging systems.Traditional reconstruction algorithms for Fourier ptychographic microscopy imaging require multiple iterations,need to increase the redundancy of low-resolution image data,have high computational complexity,and have poor reconstruction performance.In this paper,a deep convolutional neural network is used for the study of Fourier stacked microscopy reconstruction method.The innovation of the paper mainly lies in the proposed Fourier stacked microscopic imaging reconstruction method with migrating multi-convolutional network feature fusion,which improves the reconstruction quality and reconstruction speed of reconstructed images,reduces the redundancy and computational complexity of low-resolution image data,and is robust to the added noisy images.The specific work of this paper is as follows:(1)The paper utilizes the characteristics of the residual network to improve the information utilization,and uses the migrated residual network for the Fourier stacked microscopy reconstruction algorithm.The migrated residual network model is constructed and the network model is trained to reconstruct the image end-to-end from the input image data to reduce the computational complexity of the reconstructed image.To facilitate the network training,large-scale simulation and small-scale real data are used to construct the data set during the network model training,and the data set is constructed with the data synthesis input method to reduce the number of channels of input data.It is experimentally verified that the Fourier stacked microscopy reconstruction algorithm with migrated residual network has higher reconstruction quality,lower computational complexity,and shorter reconstruction time than the traditional phase recovery algorithm,and the migrated residual network model is stable and feasible.(2)In order to further improve the performance of the reconstruction method,enhance the generalization ability and its robustness,the Fourier stacked microscopic imaging reconstruction method of migrating multi-convolutional network feature fusion is proposed.In the image reconstruction process,the migrating multi-convolutional network structure is built,and the image features extracted by different convolutional models are complementary to realize the multi-convolutional network feature cascade fusion,and the front-end upsampling reconstruction network reconstructs the image based on the extracted fused features to enhance the texture details of the reconstructed result image.The experimental results show that the multi-convolutional network model is stable and feasible,achieves image reconstruction in a relatively short time,is robust to Gaussian noise,and has a good reconstruction effect under the verification of real equipment acquisition images.