高分一号卫星4种融合方法评价

以国产高分辨率卫星高分一号作为数据源,应用Pansharp融合、HPF融合、Gram-Schmidt融合和SFIM融合4种方法对高分一号卫星2m全色及8m多光谱数据进行了融合试验,并对融合结果的空间信息融入度和光谱保真度进行了评价。选取标准差、熵及联合熵、平均梯度、相对偏差4种客观评价指标对融合结果进行了计算与分析。研究结果表明:对于高分一号卫星,4种方法均显著提高了影像的空间分辨率,同时较好地保留了影像的光谱信息,提高了影像的利用率。其中Pansharp融合综合表现最好,HPF方法边界最为清晰,SFIM方法的光谱保真度最高,GramSchmidt融合在近红外波段效果最好。根据不同的研究目的,使用适宜的融合方法及参数,可以使高分一号卫星影像更好地为生产及科研工作服务。

Evaluation Study of Four Fusion Methods of GF-1 PAN and Multi\|spectral Images

The improve fusion algorithms of Pansharp,HPF,Gram\|Schmid and SFIM were studied in this paper based on the pan and multi\|spectral images of GF\|1 in Beijing Capital International Airport and the area around.We evaluated the gains of spatial information and the fidelity of spectral information using the indicators of the standard deviation,the entropy and joint entropy,the mean grads and the relative deviation.The results showed that all of the four evaluation indicators can improve the spatial information and keep the spectral information of images as well for GF\|1 images.The utilization of the images was improved.The comprehensive effect of Pansharp was the best.The method of HPF obtained the clearest boundaries.The result had the highest fidelity of spectral information which was calculated using the method of SFIM.Gram\|Schmidt represented the best in the near\|infrared band fusion.The study results indicates that GF-1 images will serve the production and science research better if we use different fusion algorithms and parameters according to different study purposes.