一种基于局部方差相似度的自适应图像融合算法

An adaptive image fusion algorithm based on local variance similarity

针对现有多光谱和全色图像融合算法空间和光谱特性难以兼顾的问题,文章提出了一种基于局部方差相似度的自适应图像融合算法,通过局部方差相似度自适应选择融合规则来改善现有问题。首先对多光谱图像主成分分析(principal component analysis,PCA)变换后的第一主分量和全色图像进行小波变换;其次根据系数矩阵局部方差相似度在2选1和加权平均之间自适应选择小波系数融合策略;最终由对应的逆变换获取融合图像。实验采用Landsat7和QuickBird卫星数据进行算法验证,并与基于PCA变换、小波变换(wavelet transform,WT)、基于局部算法的改进小波算法和自适应IHS(intensity hue saturation)算法进行比较分析。实验结果表明,该方法在提高融合图像空间和光谱质量上,综合性能优越。

Given that current multispectral and panchromatic image fusion methods are unable to bal- ance the spatial and spectral characteristics, an adaptive image fusion algorithm based on local variance similarity is proposed, which can adaptively select fusion strategy based on local variance similarity to improve the existing problem. Firstly, the first principal component of multispectral image after prin- cipal component analysis（PCA） transform and panchromatic image are processed by wavelet trans- form. Then the local variance similarity of each pixel point of wavelet coefficient matrix is calculated. Secondly, the fusion strategy is adaptively selected between alternatives and weighted average accord- ing to similarity. Finally, the fusion result is obtained by inverse transformation. The experiments are conducted by using Landsat7 and QuickBird satellite data, and a comparison with PCA transform based fusion algorithm, wavelet transform（WT） based fusion algorithm, new wavelet method based on local algorithm and adaptive intensity hue saturation（IHS） fusion algorithm is made. The experi- mental results show that the comprehensive performance of the proposed method in improving spatial and spectral quality of fusion image is better.