程传阳, 王忠华. 基于波段背景清晰度的小波变换高光谱遥感图像融合[J]. 南昌航空大学学报(自然科学版), 2018, 32(2): 68-75. DOI: 10.3969/j.issn.1001-4926.2018.02.010
引用本文: 程传阳, 王忠华. 基于波段背景清晰度的小波变换高光谱遥感图像融合[J]. 南昌航空大学学报(自然科学版), 2018, 32(2): 68-75. DOI: 10.3969/j.issn.1001-4926.2018.02.010
CHENG Chuan-yang, WANG Zhong-hua. Wavelet Transform-based Hyperspectral Remote Sensing Image Fusion with Band Background Clarity[J]. Journal of nanchang hangkong university(Natural science edition), 2018, 32(2): 68-75. DOI: 10.3969/j.issn.1001-4926.2018.02.010
Citation: CHENG Chuan-yang, WANG Zhong-hua. Wavelet Transform-based Hyperspectral Remote Sensing Image Fusion with Band Background Clarity[J]. Journal of nanchang hangkong university(Natural science edition), 2018, 32(2): 68-75. DOI: 10.3969/j.issn.1001-4926.2018.02.010

基于波段背景清晰度的小波变换高光谱遥感图像融合

Wavelet Transform-based Hyperspectral Remote Sensing Image Fusion with Band Background Clarity

  • 摘要: 高光谱遥感图像具有大量的光谱波段,有助于地物的精细分类与识别。然而,随着波段的增加,数据冗余度也相应增加,使得图像融合计算量增大,过程繁杂。因此,提出了基于波段背景清晰度的小波加权平均高光谱图像融合方法:以J-M距离和最佳指数值为指标提取优选波段组合,以减少波段数据冗余,提高信息互补,使之有利于高光谱图像融合。该算法包含如下3个步骤:首先,利用J-M距离和最佳指数选择原则,从HSI高光谱遥感图像的115个波段中提取所需优选波段及优选波段组合。其次,采取单波段遥感图像背景清晰度处理的EM算法对所选波段进行遥感图像增强预处理。最后,采用小波加权平均的像素级融合优选波段遥感图像增强数据,使得融合图像质量更优。实验结果表明:本文方法提高了融合图像的标准差、信息量和清晰度,使地物空间细节能力增强,地物特征更加明显。

     

    Abstract: Hyperspectral remote sensing images have a large number of spectral bands, which contribute to the fine classification and recognition of ground objects. However, with the band increasing, the data redundancy is raised correspondingly, making the image fusion computation complicated and the process complicated. Therefore, in this paper, combining the band background clarity, a wavelet weighted average method is proposed for hyperspectral image fusion. Taking the J-M distance and the best index value as the index, the optimal band combination is extracted to reduce the band data redundancy and improve the information complementarily, which is propitious to hyperspectral image fusion. The algorithm contains the following three steps. Firstly, using J-M distance and the principle of the best index selection, the optimal band and the preferred band combination are extracted from 115 bands of HSI hyperspectral remote sensing images. Secondly, a single band remote sensing image background clarity processing EM algorithm for the remote sensing image are used to enhance the pretreatment of the selected band. Finally, the pixel level of wavelet weighted average is used to optimize the enhanced data of the band remote sensing image, which makes the quality of the fused image better. The experimental results show that this method improves the standard deviation, information content and clarity index of the fused image, which enhances the spatial detail and the surface features.

     

/

返回文章
返回