一种基于多尺度低秩分解的红外与可见光图像融合方法

Infrared and Visible Image Fusion Method Based on Multiscale Low-Rank Decomposition

针对传统红外与可见光图像融合方法存在融合图像场景对比度低、热红外目标不明显、细节与纹理不够清晰的问题,提出了一种基于多尺度低秩分解的红外图像与可见光图像融合方法。该方法通过多尺度低秩分解,将红外图像、可见光图像分别分解为多层次局部低秩图(显著图)和全局低秩图;充分考虑分解图像特点,有针对性地设计最优融合规则,有效融合红外和可见光图像的互补信息;基于融合规则重构融合图像。基于公开数据集进行实验验证。结果表明,本文所提方法能够生成目标清晰、细节丰富的融合图像,相比于其他方法,具有更好的视觉效果和融合精度。

Traditional methods of infrared and visible image fusion generally possess disadvantages of low contrast,inconspicuous thermal infrared target,and insufficient details and textures.To address these problems,an infrared and visible image fusion method based on multiscale low-rank decomposition was proposed in this study.First,multiscale low-rank decomposition was used to decompose the infrared and visible images into multilevel local parts(saliency parts)and global low-rank parts,respectively.Second,optimal fusion rules were designed to effectively integrate the complementary information of infrared and visible images by comprehensively analyzing the characteristics of decomposed images.Finally,the fusion of the images was reconstructed according to the proposed fusion rules.The proposed fusion method was tested and verified using an open dataset.Experimental results show that the proposed method can obtain fusion images with clear targets and rich details.Further,it produced an enhanced visual effect and higher accuracy compared with other state-of-the-art fusion methods.