为了提高多光谱图像与全色图像的融合质量, 研究了多种滤波器和融合算法, 提出了基于多级引导滤波器的区域融合方法。采用该方法对多光谱图像进行插值,利用改进的分水岭算法对全色图像进行区域划分, 并将划分结果映射至每个多光谱图像, 然后将多光谱图像与全色图像利用多级引导滤波器分别进行滤波, 得到各自的细节信息, 最后根据每个区域中全色图像和多光谱图像的关系指标局部相关系数与4阶相关系数的大小, 对细节信息进行区域融合, 得到融合后的多光谱图像。结果表明, 该算法充分保留了多光谱图像的光谱信息, 并尽可能多地注入了全色图像的细节信息, 成功地提高了多光谱图像的融合效果。

In order to improve the fusion quality of multi-spectral images and panchromatic images, after studying a variety of filters and fusion algorithms, a regional fusion method was proposed based on multistage guide filters. First of all, the new method was adopted to interpolate the multispectral image; then the improved watershed algorithm was used to divide the panchromatic image, and the result was mapped to each multispectral image. Secondly, a multistage guide filter was used to filter the multispectral image and the panchromatic image, and their details were obtained respectively. Finally, according to the local correlation coefficient and the 4-order correlation coefficient between the panchromatic image and multispectral image, the detailed information was fused and the fusion images were obtained. The results show that the algorithm retains the multispectral image spectrum information and at the same time, details of the panchromatic image were put into the fusion image as much as possible. The method improves the effect of multispectral image fusion.