基于卷积神经网络融合的彩色傅里叶叠层显微重建  

Color Fourier Ptychographic Microscopic Reconstruction Based on Convolutional Neural Network Fusion

作　　者：李杰[1] 王浩明 LI Jie;WANG Haoming(College of Electronic Information Engineering,Changchun University,Changchun 130022,China)

机构地区：[1]长春大学电子信息工程学院,长春130022

出　　处：《长春大学学报》2024年第2期1-7,共7页Journal of Changchun University

基　　金：吉林省科技厅项目(YDZJ202301ZYTS180)。

摘　　要：针对彩色傅立叶叠层显微重建存在图像获取时间长、采集低分辨率图像数量多等问题,采用基于卷积神经网络的图像融合方法来实现彩色傅立叶叠层显微重建。该方法基于图像融合原理,将单通道低分辨率图像重建的灰度高分辨率图像与相同视场下的彩色低分辨率图像融合,成功地重建彩色FPM图像。在保证图像恢复质量的同时减少了2/3的采集时间。实验结果表明,所提算法可以获得色彩不失真的彩色FPM图像,定量评价指标方均根误差小于0.01,结构相似性参数大于0.89。In view of the problems of long image acquisition time and large number of low-resolution images in collecting color Fourier ptychographic microscopic reconstruction,this paper adopts an image fusion method based on convolutional neural network to realize color Fourier ptychographic microscopic reconstruction,which fuses the gray-scale high-resolution image reconstructed by single-channel low-resolution image with the color low-resolution image in the same field of view based on the principle of image fusion,and successfully reconstructs the color FPM image.This method reduces the acquisition time by 2/3 while ensuring the quality of image restoration.The experimental results show that the proposed algorithm can obtain color FPM images without color distortion,and the root mean square error of the quantitative evaluation index is less than 0.01 and the structural similarity parameter is larger than 0.89.

关 键 词：傅里叶叠层显微重建 彩色图像重建 图像融合 卷积神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]