采用冗余字典稀疏表达的红外与水汽云图融合

卫星云图作为典型的多光谱遥感图像,因各个遥感器成像波段的差异,致使云图间既有一定的相关性,又存在一定的差异,故可认为云图包含2种特征:共性特征和个性特征.一种稀疏表示的云图融合方法,能够把多幅云图在一个过完备字典上进行稀疏表示,使用稀疏系数作为云图的特征,然后对不同图像的个性特征根据稀疏系数向量的1范数决定权重因子,融合云图可以由共性特征和融合后的个性特征联合表示.实验表明,该方法的融合云图无论在客观指标还是视觉效果上都优于传统方法,蕴藏了更为丰富的天气信息.

Infrared and Water Vapor Cloud Image Fusion Using Redundant Dictionary Sparse Representation

In this paper, a method of the cloud image fusion based on sparse representation is presented. For the typical multi-spectral remote sensing images, there is not only a certain correlation between clouds, but also a certain difference due to the varying sensor imaging spectrum. So in this research, we consider two characteristics with each cloud, that is, common features and individual traits. First we sparsely decompose the cloud image according to a complete dictionary and use the sparse coefficients as the characteristic of the cloud. Then, we use 1 norm of sparse coefficient vectors belonging to different cloud images to determine weighting factor for the individual traits. The fused cloud image can be expressed by the common features and the fused individual traits. Experiments show that the fused cloud image based on the sparse representation is better than the traditional method in both the objective indicators and visual effect, and contain more helpful weather information.