传统的可见光与红外稀疏表示融合方法, 采用图像块构造解析字典或者学习字典, 利用字典的原子表征图像的显著特征。这类方法存在两个问题, 一是没有考虑图像块与块之间的联系, 二是字典的适应能力不够并且复杂度高。针对这两个问题, 本文提出可见光与红外图像结构组双稀疏融合方法。该方法首先利用图像的非局部相似性, 将图像块构建成图像相似结构组, 然后对图像相似结构组进行字典训练, 采用双稀疏分解模型, 有效结合解析字典和学习字典的优势, 降低了字典训练的复杂度, 得到的结构字典更加灵活, 适应性提高。该方法能够有效提高红外与可见光融合图像的视觉效果, 经对比实验分析, 在主观和客观评价上都优于传统的稀疏表示融合方法。

In the traditional visible and infrared image fusion based on sparse representation, the analytical and learning dictionaries are constructed by using image blocks, and the atoms of the dictionaries are used to represent the salient features of the image. This method creates two problems. First, the relationships among the patches are ignored. Second, the dictionaries have poor adaptability and are complicated to learn. Aiming at solving these two problems, a visible and infrared image fusion method based on a structured group and double sparsity is proposed in this study. Image blocks are constructed into similarity structure groups by using the non-local similarity of the image. Then, the dictionary is built based on similarity structure groups and a double sparsity model to reduce the complexity of dictionary training, thereby improving the analytical and learning dictionaries. The obtained training dictionary is more adaptable, and the complexity of dictionary training is reduced. The experimental results demonstrate that compared with the traditional sparse representation fusion method,this method can effectively improve the visual effect of the fused image and is superior in terms of both subjective and objective evaluation.