高光谱图像融合最佳波段选择方法

Band Selection of Optimal for Hyperspectral Image Fusion

针对高光谱图像高数据维给图像处理带来的困难和影响,本文构造了高光谱图像融合的最佳波段选择新模型—联合偏度-峰度指数(Joint Skewness-Kurtosis figure,JSKF)模型,利用JSKF指数进行自适应子空间的分解和波段选择,降低高光谱数据维数;并将选择出的最佳波段组合进行了融合,实验结果表明,该方法所选择的波段信息差异较大、互补特征明显,融合后图像包含的信息量丰富,效果优于传统的自适应波段选择方法和主成分分析累计贡献率方法。

A novel band selection model called joint skewness-kurtosis figure（JSKF） is proposed to solve problem of high dimensions of hyperspectral images.The whole data base is automatically partitioned into different sub-spaces by the sign and value of JSKF.Then the optimal bands for image fusion are selected in each sub-space according to the absolute value of JSKF.This band selection algorithm is applied in OMIS hyperspectral images and the selected bands are fused by a common image fusion method.The experimental results show that the bands selected by JSKF contain richer complementary information,especially the characteristics of small targets and textures,than those selected by the conventional adaptive band selection method and cumulative contribution rate method based on principal component analysis（PCA）,and also provide improved fusion results.