光学学报, 2020, 40 (24): 2410001, 网络出版: 2020-11-23
基于深度学习的细胞骨架图像超分辨重建 
下载： 1531次
Super-Resolution Reconstruction of Cytoskeleton Image Based on Deep Learning
图 & 表
图 1. EDSR的结构。Conv表示卷积层,ResBlock表示残差模块,ReLU表示线性整流激活函数,Upsample表示上采样,Shuffle表示周期筛选
Fig. 1. Structural diagram of EDSR. Conv represents convolution layer, ResBlock represents residual module, ReLU represents linear rectification activation function, Upsample represents upsampling, and Shuffle represents cycle screening
图 2. 基于深度学习的图像超分辨重建训练过程
Fig. 2. Schematic of training process for deep-learning based image super-resolution reconstruction
图 3. 二倍降采样条件下损失函数的值与EDSR训练轮数的关系
Fig. 3. Relationship between loss function and training epoch of EDSR in the case of double down-sampling
图 4. 基于EDSR深度学习对二倍降采样细胞微管骨架实验图的超分辨重建。(a)细胞骨架图像;(b)放大图
Fig. 4. Super-resolution reconstruction of cell microtubule cytoskeleton images obtained by double down-sampling based on EDSR deep learning. (a) Images of cytoskeletons; (b) enlarged views
图 5. 基于EDSR深度学习对三、四倍降采样STORM图像的超分辨重建。(a)三倍降采样图重建图;(b)四倍降采样图重建图
Fig. 5. Super-resolution reconstruction of three and four times down-sampling STORM images based on EDSR deep learning. (a) Reconstruction of three times down-sampling images; (b) reconstruction of four times down-sampling images
表 1二倍降采样条件下各图的平均梯度值
Table1. Average gradient values of different images in the case of double down-sampling
|
表 2三倍和四倍降采样条件下各图的平均梯度值
Table2. Average gradient values of different images in the case of three and four times down-sampling
| ||||||||||||||||||||||||||||||||||||||||||
胡芬, 林洋, 侯梦迪, 胡浩丰, 潘雷霆, 刘铁根, 许京军. 基于深度学习的细胞骨架图像超分辨重建[J]. 光学学报, 2020, 40(24): 2410001. Fen Hu, Yang Lin, Mengdi Hou, Haofeng Hu, Leiting Pan, Tiegen Liu, Jingjun Xu. Super-Resolution Reconstruction of Cytoskeleton Image Based on Deep Learning[J]. Acta Optica Sinica, 2020, 40(24): 2410001.
PDF全文
