基于MTF和变分的全色与多光谱图像融合模型

Pan-sharpening Model Based on MTF and Variational Method

Pan-sharpening将高分辨率图像全色(Panchromatic,Pan)波段的空间细节注入多光谱(Multispectral,MS)波段,以生成同时具有高光谱和高空间分辨率的多光谱图像.为改善融合效果,需要考虑多光谱和全色波段的调制传输函数(Modulation transfer function,MTF).本文提出了一个新的基于MTF和变分的Pan-sharpening模型.该模型的能量泛函包括两项,第1项为细节注入项,基于高通滤波器从Pan波段中提取细节信息并注入融合图像;第2项为光谱保真项,基于MTF设计多孔小波的低通滤波器以保持MS波段的多光谱信息.在Quick Bird、IKONOS和Geo Eye数据集上的融合结果表明,该模型可以生成同时具有高空间和高光谱质量的融合图像,融合效果优于AWLP、IHS BT、HPM-CC-PSF、NAWL、快速变分等算法.

In order to provide the multispectral （MS） image with both high spectral and high spatial resolution, pan- sharpening approach introduces the spatial details of Panchromatic （Pan） band into MS band. Modulation transfer function （MTF） of MS and Pan bands is necessary for high fusion quality. This paper proposes a novel pan-sharpening model based on MTF and variational method. The energy functional of the proposed model consists of two terms. The first one is the spatial detail injection term, which injects detail information extracted from Pan band by a high-pass filter into MS image. The second one is the spectral signature preserving terms, in which a low-pass filter of ＂a trous＂ wavelet is designed to maintain the multispectral information based on MTF of MS band. The experimental results on QuickBird/IKONOS/GeoEye datasets demonstrate that this model can produce the fused MS image with high spectral and high spatial quality. The proposed model is superior to AWLP, IHS_BT, HPM-CC-PSF, NAWL and fast variational method in fusion performance.