A fast fusion scheme for infrared and visible light images in NSCT domain

• We propose a fast realization of NSCT instead of traditional NSCT for image decomposition.
• The computational efficiency is improved through combining the filters in each channel of NSCT.
• Once-for-all calculation of the global attribute is used to obtain higher efficiency.
• The fusion weights are determined by the difference between pixel and global image attribute.

Fusion of infrared and visible light images is an effective way to obtain a simultaneous visualization of details of background provided by visible light image and hiding target information provided by infrared image, which is more suitable for browsing and further processing. Two crucial components for infrared and visual light image fusion are improving its fusion performance as well as reducing its computational burden. In this paper, a novel fusion algorithm named pixel information estimation is proposed, which determines the weights by evaluating the information of pixel and is well applied in visible light and infrared image fusion with better fusion quality and lower time-consumption. Besides, a fast realization of non-subsampled contourlet transform is also proposed in this paper to improve the computational efficiency. To verify the advantage of the proposed method, this paper compares it with several popular ones in six evaluation metrics over four different image groups. Experimental results show that the proposed algorithm gets a more effective result with much less time consuming and performs well in both subjective evaluation and objective indicators.