摘要
提出了一种基于非负矩阵分解的多光谱图像与全色图像的融合算法。在非负矩阵分解过程中,将低分辨率的多光谱图像和高分辨率的全色图像作为原始数据,非负矩阵分解得到的特征基包含了原始图像的整体特征,将高分辨率图像与分解得到的特征基中的第一特征基作直方图匹配,并代替第一特征基。利用特征基进行重构,得到具有较高的空间分辨率和保持原有多光谱图像的光谱信息的融合图像。主观视觉效果分析和客观统计参数评价分析表明,基于非负矩阵分解的图像融合方法在性能上优于传统的其他融合方法。
A new method based on non-negative matrix factorization （NMF） technique is presented for multi-spectral image and panchromatic image fusion. Low resolution multi-spectral image and high resolution panchromatic image are taken as source data in NMF analysis, from which feature bases containing global features of the original images can be extracted. The first feature base is replaced by the adjusted panchromatic image, which is histogram matched to the raw first feature base. Inverse transform （reconstruction） is performed and the fusion image with higher spatial resolution and spectral component is obtained. Visual results and statistical indexes demonstrate the advantage of the proposed NMF-based image fusion method compared to common fusion approaches.
出处
《计算机工程》
CAS
CSCD
北大核心
2007年第21期169-171,共3页
Computer Engineering
关键词
图像融合
非负矩阵分解
特征基
image fusion
non-negative matrix factorization
feature base
