高分二号卫星影像融合方法比较及效果优化研究

Research on Comparison of GF-2 Image Fusion Methods and Effect Optimization

针对如何使高分二号卫星融合后影像达到最佳视觉与应用效果还没有达成统一意见的问题,本文采用Gram-Schmidt(G-S)、高通滤波法(HPF)、最邻近扩散算法(NND)、超分辨率贝叶斯法及改进算法(Pansharp及Pansharp2)等5种目前主流的遥感影像融合算法,对高分二号卫星全色及多光谱影像数据开展融合实验,梳理了影像融合质量评价体系,同时对融合结果进行通道重组实验及优化。结果表明:NND、HPF、Pansharp融合算法比较适合高分二号卫星影像,将20%近红外波段加入到绿波段时,各融合算法影像视觉效果都得到增强,其中NND融合算法光谱性良好,HPF融合算法清晰度最好,Pansharp融合算法效果稳定,可作为后备方法,实验结果将为后续高分系列卫星影像处理提供支持。

Aiming at the problem that there is no unified opinion of how to make GF - 2 satellite images to achieve the best visual and application results, this paper use the G - S, HPF, NND, Pansharp and Pansharp2 five types of current mainstream of remote sensing image fusion algorithm to carry out fusion experiments on GF - 2 panchromatic and multi - spectral image date. Combing the image fu- sion quality evaluation system, at the same time, the results of fusion image of the channel reconstruction experiments and effect opti- mization was carried out. The results showed that NND, HPF and Pansharp fusion algorithm is suitable for the GF -2, when 20% near - infrared bands added to the green band, the fusion algorithm of image visual effect have been strengthened, Among them, the NND has good spectral performance, and the HPF has the best definition. The Pansharp fusion algorithm is stable and can be used as a backup method. The experimental results will provide support for the subsequent high resolution satellite image processing.