Adaptive fusion method of visible light and infrared images based on non-subsampled shearlet transform and fast non-negative matrix factorization

The issue of visible light and infrared images fusion has been an active topic in both military and civilian areas, and a great many relevant algorithms and techniques have been developed accordingly. This paper addresses a novel adaptive approach to the above two patterns of images fusion problem, employing multi-scale geometry analysis (MGA) of non-subsampled shearlet transform (NSST) and fast non-negative matrix factorization (FNMF) together. Compared with other existing conventional MGA tools, NSST owns not only better feature-capturing capabilities, but also much lower computational complexities. As a modification version of the classic NMF model, FNMF overcomes the local optimum property inherent in NMF to a large extent. Furthermore, use of the FNMF with a less complex structure and much fewer iteration numbers required leads to the enhancement of the overall computational efficiency, which is undoubtedly meaningful and promising in so many real-time applications especially the military and medical technologies. Experimental results indicate that the proposed method is superior to other current popular ones in both aspects of subjective visual and objective performance.