摘要
提出一种利用BP神经网络进行三维宽场显微图像复原的非线性映射方法,将三维图像转化为二维图像进行处理,利用神经网络的学习能力,通过训练,建立含有散焦信息的二维模糊图像与二维清晰图像之间的映射关系,然后对切片堆叠进行逐幅复原,从而实现显微图像的三维复原·得到的复原图像在视觉上和定量分析上都获得了很好的效果·由于采用小规模神经网络,训练时间短,计算量小,使实时复原成为可能·
The restoration for 3D （three dimensions） image of wide-field microscope needs to process very large data size, and it spend much time. In this paper, a new method is proposed for 3D image restoration of wide-field microscope based on the BP neural network. The initial step of the method is to transform a 3D image into a series of 2D images. Then, the mapping relationship between the 2D blurring image with defocusing message and 2D clear image is established by training the BP neural network,which has a high ability of learning. Following, every 2D section image of the stack is restored in succession. As a result, the restoration of 3D image of wide-field microscope is achieved. Extensive tests demonstrate that this method has a satisfying restoration effect both in visual impression and quantitative analysis. For adopting small dimensional neural network, the training time is little, and the operational amount is small. Thus, it is possible to realize the real time restoration.
出处
《光子学报》
EI
CAS
CSCD
北大核心
2006年第3期473-476,共4页
Acta Photonica Sinica
基金
高等学校博士学科点专项科研基金(20020007006)资助项目
关键词
图像复原
神经网络
三维宽场显微图像
非线性映射
Image restoration
Neural network
3D wide-field microscope image
Non-linearity mapping
