摘要
依据光学信息论中的多维随机变量的联合熵、条件熵、平均互信息之间关系和互信息的链式法则,提出了面向挤出中聚合物形态的层析图像处理方法。它包括层析图像中的背景与分离、目标标度变换、亚像素插补的图像预处理和多模式目标特征图像的提取和测量。实验表明,通过该方法不仅能将目标形态从目标与背景之比为1∶9的层析图像中分辨出来,而且能提高目标图像的空间分辨率30%,解决了由于目标图像信息量小、对比度差而使得层析图像测量方法难以在工程中应用的问题。
A novel method of tomography image processing is presented for polymer configuration of extrusion,according to the relation of joint entropy,condition entropy,mean mutual information and mutual information chain rule for multidimension random variable in the theory of optics information.It includes the pretreatment of the tomography image segmentation of background and object,grey level transformation,subpixel interpolation and the object character region extract and measurement with multimode.From the experimental result,it can clearly be seen that these methods can not only distinguish object's configuration from the tomography image in which the ratio of pixel between background and object in tomography image is about 9 to 1,but also increase spatial resolution 30 percent,and solve the problem of the application of the tomography image in measurement engineering for little information and low contrast in object tomography image.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2005年第z1期212-216,共5页
Optics and Precision Engineering
基金
国家自然科学基金项目(No.50373012
No.29874015)
广东省自然科学基金研究团队项目(110D40200)