摘要
简要介绍了盲源分离技术和独立成分分析的基本思想。阐述和讨论了核函数和核独立成分分析(KI-CA)的基本原理,详细介绍了基于核典型相关性分析的核独立成分分析的基本算法。用KICA对一维混合信号的分离进行了模拟实验,目的是验证KICA的优越性能。之后通过实验分别讲述了KICA在自然图像和遥感影像处理的应用。最后指出了ICA所固有的分离结果的序号和幅度不确定性的问题。实验表明,KICA能够很好地分离混合图像,而且经处理的遥感影像能够更清晰地反映地表情况。
Basic principles of blind source separation technology and independent component analysis were presented. Principles of kernel functions and kernel independent component analysis were discussed. Furthermore, the algorithm of KICA based on kernel canonical correlation analysis was introduced in details. Simulation on one-dimensional mixed signals was presented to verify the superiority of KICA. The problem of ICA separation inherent uncertainty of the order and the magnitude of the results was indicated finally. Later experiments on natural images and remotely sensed images show that KICA can separate mixed images successfully, In addition, processed images by this approach can better reflect the land surface conditions.
出处
《计算机应用研究》
CSCD
北大核心
2008年第1期297-299,共3页
Application Research of Computers
基金
国家自然科学基金资助项目(40371078)
科技部国际合作重点项目(2004DFA06300)
长江学者和创新团队发展计划资助项目
博士后基金资助项目(2005037317)
关键词
核独立成分分析
图像处理
遥感影像
kernel independent component analysis （KICA）
image processing
remotely sensed image
