一种基于à trous小波和联合稀疏表示的遥感图像融合方法

采用训练字典的稀疏表示方法能反映信号的本质特征和内在结构。针对遥感多光谱图像和全色图像融合存在的光谱失真问题,提出了一种基于àtrous小波和联合稀疏表示的融合方法。首先对多光谱图像进行IHS变换,然后对全色图像和变换后的多光谱亮度分量进行àtrous小波变换,对其低频分量进行字典训练,采用联合稀疏表示模型进行分解得到公共成分和独特成分,最后对稀疏系数进行融合。通过对山区和城区不同场景的IKONOS遥感数据进行实验,融合结果不仅在空间分辨率得到了提高,并且光谱分辨率保持较好,目视判读和量化分析表明其多数性能优于目前常用的传统算法。

A Method on Remote Sensing Image Fusion based on à trous Wavelet Transform and Joint Sparse Representation

Sparse representation using the trained dictionary can reflect the inherent characteristics and structure of signals.A novel fusion method based on the à trous wavelet transform and joint sparse representation for multi\|spectral image and panchromatic image is proposed,aiming to solve the spectral distortion.Firstly,the IHS transform is applied to multi\|spectral image.Then,the panchromatic image and the intensity components of multi\|spectral image are decomposed by à trous wavelet transform.The trained dictionary is learned from their low components.By exploiting joint sparse representation model on their low frequency components,common component and innovation component can be obtained.The finally result is obtained by fusing the sparse coefficients.Experimental results on urban area and mountainous area from IKONOS satellite indicate that the fused image has higher spatial resolution and great spectral fidelity.And the proposed method outperforms traditional methods by visual analysis and quantitative evaluation.