基于非负最小二乘法的全色与高光谱图像融合

Panchromatic and Hyperspectral Images Fusion Based on Non-negative Least Squares Algorithm

现有光学遥感图像融合方法主要针对全色与多光谱图像,直接将其用于全色与高光谱图像融合存在以下问题:高光谱图像波段数量远多于多光谱图像,通过高光谱波段加权合成低分辨率全色图像,容易出现灰度失真;高光谱图像与全色图像的空间分辨率相差很大,采用现有的加性变换融合方法,会导致融合图像中部分地物出现光谱或细节失真。为此,文章提出了基于非负最小二乘法的全色与高光谱图像融合方法。首先对高光谱图像进行波段压缩,得到波段压缩的高光谱图像;然后对波段压缩的高光谱图像及全色图像进行非负最小二乘拟合,得到低分辨率全色图像;最后,采用比值变换模型生成融合图像。试验表明该方法的光谱与细节保真效果好,优于对比方法。

The existing optical remote sensing image fusion methods focus on panchromatic and multispectral images. There will be the following problems while directly using these methods to deal with panchromatic and hyperspectral images. Due to the fact that hyperspectral bands are far more than multispectral bands, the existing methods are prone to grayscale distortion while synthesizing low-resolution panchromatic images by the weighted summation of hyperspectral bands. On the other hand, the existing optical remote sensing image fusion methods mainly use additive transformation, which can cause spectral or detail distortion in some ground objects because there are great spatial resolution difference between hyperspectral and panchromatic images. To solve the problems, this paper proposes the non-negative least squares algorithm based panchromatic and hyperspectral images fusion method. Firstly, this method reduces the bands of hyperspectral images by band compression. Then the low-resolution panchromatic images are generated by orthogonal least squares algorithm from the reduced bands and the panchromatic images. Finally, the fusion images are obtained by a ratio transformation model. The experiments demonstrated that the proposed method had a good performance on fusion quality, and was superior to the existed methods.