小波多分辨分量相关性图像融合方法

针对多光谱图像与高空间分辨率全色图像的融合,提出了一种基于小波多分辨分量相关性的图像融合方法.该方法在原始图像小波分解的基础上,对分解得到的高低频分量采用不同的融合规则进行融合,最后逆变换重构得到最终的融合图像.其中,对分解得到的低频分量采用局域相关系数加权融合计算,对高频分量设置一定的约束条件,使相关性较强的部分得到增强,而相关性较弱的部分继续保持.通过实验,与几种典型的图像融合方法就相关系数和光谱扭曲度等指标进行对比,验证该方法在光谱信息保持和空间分辨率提高方面的优势,并实验分析了不同相关系数门限值T的选取对最终融合效果的影响,以针对不同的应用场景做适当的选取.

Wavelet multi resolutionweight correlationimage fusion method

An image fusion method which based on wavelet multi resolution of weight correlation was introduced for the fusion of multispectral image and panchromatic image. On basis of wavelet decomposition for original image, the method fused the high and low weights image with different fusing rules, and the final fused image was obtained by taking inverse wavelet transform. The low frequencies were fused by weighted calculating with the coefficient of local correlation, and a certain constraints for the high frequencies were defined, so that part of the strong correlation between strengthened, and the weak part preserved. In the end, the experiment indicates its advantage to preserve spectra information and improve spatial resolution by comparing the correlative coefficient and the spectral distortion with several typical image fusion methods. The effect of fusion for different threshold value of T was also analyzed by experiment, so that appropriate value can be set to adapt different scene.