基于RBF神经网络的COSM图像复原算法  

Restoration method for COSM image based on RBF neural network

作　　者：贺可鑫[1] 何小海[1] 陶青川[1] 王宇[1] 

机构地区：[1]四川大学电子信息学院,成都610064

出　　处：《计算机应用》2009年第1期78-80,85,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60372079);教育部科学技术研究重点项目(107094)

摘　　要：在计算光学切片显微技术成像中,每幅切片图像都要受到其他离焦层信息的干扰,引起图像模糊。针对此问题提出了一种基于RBF神经网络的复原算法,利用神经网络的学习和泛化能力,用一组样本图像对网络进行训练,建立含有离焦模糊信息的模糊三维图像与其对应清晰图像间的非线性映射关系,然后用训练好的网络进行图像复原。实验证明该算法的复原速度快,且复原的三维图像在主观视觉和定量分析上都获得了较好的效果。In the process of obtaining 3D images by Computational Optical Sectioning Microscopy method （ COSM）, every slice image is disturbed by other defocusing messages and the 3D images are blurred. In order to resolve this problem, a new restoration method based on the RBF neural network was proposed. The nonlinear mapping relationships between the 3D blurred images with defocusing messages and 3D clear images were established by training the Radial Basis Function （RBF） neural network that has the ability of learning and generalizing with a group of COSM images. Then 3D images that need restoring could be restored by the trained neural network. Experiment demonstrates that the speed of this method is high and this method has satisfying restoration performance in both visual impression and quantitative analysis.

关 键 词：图像复原 神经网络 计算光学切片显微技术 非线性映射 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]