自适应高斯滤波与SFIM模型相结合的全色多光谱影像融合方法

Panchromatic and Multi-spectral Fusion Method Combined with Adaptive Gaussian Filter and SFIM Model

全色-多光谱影像融合技术可以显著提高遥感影像的地物判别能力,但是空间信息融入度与光谱信息保真度是相互矛盾的一组性质,一般方法往往无法平衡这两方面。SFIM算法具有良好的光谱信息保持能力,但是其空间信息融入度较差,影响了整体的融合效果。为此,本文分析了SFIM模型的原理与特点,提出一种自适应高斯滤波与SFIM模型相结合的全色多光谱影像融合方法(AGSFIM)。以均值调整后的多光谱整体平均梯度为标准来计算高斯滤波的最优参数,将下采样全色影像的清晰度调整至同样水平,以保证融合结果的空间信息融入度与光谱信息保真度之间的平衡。利用6种融合算法对"北京二号"(Beijing-2)、"资源三号"02星(ZY-3 02)数据进行对比试验,表明在良好的光谱保持能力的前提下,改进方法可以有效克服SFIM算法空间信息融入不足的缺点。

Panchromatic and multi-spectral fusion technology can increase feature discriminant ability of remote sensing images.However,the abilities of fusing spatial information and keeping spectral information are conflict,and are hard to be balanced by common algorithms.SFIM（smoothing filter-based intensity modulation）can keep spectral information effectively,but is difficult to fuse spatial information which will reduces the holistic effect.Pointing to this problem,this paper analyzes the principles and characters of SFIM model,and proposes a fusion method combined with adaptive Gaussian filter and SFIM model（AGSFIM）.Computing optimal parameter of Gaussian filter based on entirety mean-value-adjusted average gradient of multi-spectral bands,and adjusting down-sampled panchromatic image to same sharpness level which can confirm the balances of spatial information fusing ability and spectral information keeping ability.Beijing-2 and ZY-3 02 data are applied to test and six different fusion methods are used to compare.The experiments show that AGSFIM can effectively overcome SFIM’s shortage and increase fusion images’ spatial information.