一种新的全色与多光谱图像融合变分模型  

A New Variational Model for Panchromatic and Multispectral Image Fusion

作　　者：马宁[1,2] 周则明[2] 张鹏[2] 罗立民[1] 

机构地区：[1]东南大学影像科学与技术实验室,南京210096 [2]解放军理工大学气象学院,南京211101

出　　处：《自动化学报》2013年第2期179-187,共9页Acta Automatica Sinica

基　　金：国家自然科学基金(41174164;41005018)资助~~

摘　　要：图像融合是提供包含各输入图像互补信息的单幅图像的有力工具.本文提出了一种新的用于全色和多光谱图像融合的变分模型.在Socolinsky对比度模型的基础上构造了一个改进的能量泛函最小化问题,以寻找最接近全色图像梯度的解.为了提高多光谱图像的空间分辨率,并尽可能地保持其原有的光谱信息,还将光谱一致项、波段间相关项和对比度增强项引入融合模型.在IKONOS和QuickBird数据集上测试了该模型的性能.实验结果表明该模型可以生成同时具有高空间质量和高光谱质量的融合图像.Image fusion is a powerful tool to provide a single image which contains the complementary information from the input images. A novel variational model is presented for panchromatic and multispectral image fusion. Based on the Socolinsky＇s contrast model, an advanced energy minimization problem is posed to find the solution whose gradient is closest to that of the panchromatic image. To improve the multispectral image＇s spatial resolution and preserve its spectral information as much as possible, spectral coherence, interband correlation and contrast enhancement terms are explicitly enforced into the fusion process. The performance of our model is evaluated on several IKONOS and QuickBird datasets. Experimental results show that our model can produce images with both high spatial and high spectral quality.

关 键 词：图像融合 变分 能量泛函 光谱一致性 IKONOS QUICKBIRD 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]