基于边缘增强与光谱特性保持的Pan-sharpening融合模型

为生成兼具高光谱质量与高空间质量的融合图像，本文提出了一种新的Pan-sharpening变分融合模型.通过拟合退化后的全色（Panchromatic，Pan）波段图像与低分辨率多光谱（Multispectral，MS）波段图像间的线性关系得到各波段MS图像的权重系数，计算从Pan图像抽取的空间细节；基于全色波段图像的梯度定义加权函数，增强了图像的强梯度边缘并对因噪声而引入的虚假边缘进行了抑制，有效地保持了全色波段图像中目标的几何结构；基于MS波段传感器的调制传输函数定义低通滤波器，自适应地限制注入空间细节的数量，显著降低了融合MS图像的光谱失真；针对Pan-sharpening模型的不适定性问题，引入L1正则化能量项，保证了数值解的稳定性.采用Split Bregman数值方法求解能量泛函的最优解，提高了算法的计算效率.QuickBird、IKONOS和GeoEye-1数据集上的实验结果表明，模型的综合融合性能优于MTF-CON、AWLP、SparseFI、TVR和MTF-Variational等算法.

Pan-sharpening Model on Account of Edge Enhancement and Spectral Signature Preservation

In order to provide a multispectral (MS) image with both high spectral and spatial qualities, a novel pan-sharpening model is proposed based on the variational method. The weight coefficients of MS bands are obtained by linear regression of the degraded panchromatic (Pan) and the original MS images. After that the spatial details are extracted from the Pan image and are injected into the MS image. The weight function is defined to enhance the strong edges with the gradient of the Pan image and suppress the false edges caused by image noise, so as to reserve the geometrics structure of the Pan image effectively. A low-pass filter is developed with the modulation transfer function (MTF) of the multispectral band sensors, which restrains the number of spatial details merged into MS images adaptively and reduces the spectral distortion of fused MS images. To deal with the ill-posed problem formulated by fusion operation, the L1 regularization term is introduced into the variational framework to ensure the stability of the numerical solution. The split Bregman method, which can improve computational efficiency, is used to acquire the optimization solution of the energy functional. The experimental results on QuickBird/IKONOS/GeoEye-1 datasets demonstrate that the proposed model can achieve competitive fusion performances in comparison with MTF-CON, AWLP, SparseFI, TVR and MTF-Variational methods.