摘要
为了对工业CT(ICT)图像中的线特征进行有效提取,提出一种基于二阶Wedgelet快速分解的特征提取方法。首先,介绍了二阶Wedgelet分解的定义,根据二阶Wedgelet基函数的构成将其划分为3种(同边型、邻边型、对边型)不同类型。根据不同类型基函数间的空间关系设计出二阶Wedgelet快速分解算法。在此基础上,在不同的单尺度上分别提取了各自的候选线特征。最后,结合多尺度四叉树结构,融合不同单尺度的候选线特征得到最终提取结果。对实际ICT图像进行了线特征提取实验,对比试验结果验证了提出方法的有效性和先进性,表明基于二阶Wedgelet快速分解的ICT图像线特征提取方法能有效地提取ICT图像感兴趣的线特征,对其它类型图像的线特征提取也具有一定的参考价值。
To effectively extract the line features of an industrial CT（ICT）image,a feature extraction method based on second order Wedgelet decomposition was presented.Firstly,the definition of the Wedgelet decomposition was introduced.According to the different constitutions of second-order Wedgelet basis functions,they were divided into three kinds（same side type,neighbouring type and apart type）.Depending on the spatial relationships of different kinds of second order Wedgelet basis functions,the second order Wedgelet rapid decomposition algorithm was designed.Then,based on the proposed algorithm,candidate line features at different mono-scales were extracted respectively.Finally,considering the multiscale quadtree structure,the line features were fused to get the final extraction result.The actual ICT images were used to an experiment,comparative test results show the feasibility and advance of the line features extraction method based on second order Wedgelet rapiddecomposition.For applying to the ICT images,the method not only extracts effectively interest line features,but also lays a solid foundation for the feature description of the other types images.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2015年第11期3238-3245,共8页
Optics and Precision Engineering
基金
重庆市教委科学技术研究项目(No.KJ131505)
重庆市基础与前沿研究计划资助项目(No.cstc2014jcyjA40040)
