计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (3): 71-71.

• 学术探讨 • 上一篇    下一篇

一种基于特征量积的遥感图像融合方法

何贵青 郝重阳   

  1. 西北工业大学 西北工业大学电子与信息工程研究所
  • 收稿日期:2006-02-22 修回日期:1900-01-01 出版日期:2007-01-21 发布日期:2007-01-21
  • 通讯作者: 何贵青

A Feature Product-Based Remotely Sensed Image Fusion Method

GuiQing He   

  • Received:2006-02-22 Revised:1900-01-01 Online:2007-01-21 Published:2007-01-21
  • Contact: GuiQing He

摘要: 本文提出一种基于特征量积的遥感图像融合方法。以目前卫星遥感领域最常见的传感器图像多光谱图像和全色图像为例,首先对多光谱图像作IHS变换,并对全色图像和多光谱图像的亮度分量I进行小波分解,得到不同分辨率下的小波系数特征,然后根据特征量积判决准则对多光谱图像的I分量与全色图像进行特征融合,并对高频域融合规则进行一致性检测,进而用融合结果替代多光谱图像的亮度分量I,最后,作IHS反变换得到融合图像。实验结果表明,融合图像既显著地融入了高空间分辨率图像的细节信息,又有效地保留了多光谱图像的光谱信息,同时信息丰富程度增加,从而使融合图像在保持光谱信息和提高空间分辨率两个方面的综合性能达到很好的平衡。

关键词: 图像融合, 小波变换, 小波系数特征, 特征量积

Abstract: In this paper, a feature product -based image fusion method between multi-spectral image and PAN image is proposed. Firstly, the multi-spectral image is transformed by IHS transform, and the components of I, H and S are obtained. Secondly, panchromatic images and the component I of the multi-spectral image are merged with Feature Product -based fusion method, furthermore we verify the consistency of the wavelet coefficient feature,and I component is replaced by the merged data. Finally, the fusion image is obtained by inverse IHS transform. The experimental result indicates that the new method is effective, and it enhances spatial detail information of the fusion image largely and preserves spectral information of the original Multi-spectral image well.

Key words: image fusion, wavelet transform, wavelet coefficient feature, feature product