Fusion of NSCT infrared and visible images based on multi-judgment and WLS optimization

In order to overcome some defects of traditional methods and the insufficiency of single feature extraction information, while further improving the fusion of infrared and visible images, a method with strong adaptability to different types of features is sought. A non-subsampled contourlet transform (NSCT) infrared-visible image fusion method based on multi-judgment and weighted least squares optimization (WLS) is proposed. Firstly, NSCT is used to decompose the image at multiple scales to obtain the low-frequency and high-frequency subbands of the image. Secondly, the low-frequency sub-band selects local squared entropy and sum-modified laplacian (SML) to complement each other, which extracts a small amount of detailed information under guaranteed good contrast. The high-frequency subbands fully considers the importance of the underlying features, and selects phase consistency (PC), the local weighted sum-modified Laplacian (WSML) and the local weighted energy (WLE) to complement each other to fuse the detail layer. They are optimized by WLS, and the fused image details are more natural and more suitable for human visual perception. Finally, the inverse transform is performed on the fused low-frequency and high -frequency subbands to obtain a fused image. Through the experimental verification of images with different types of characteristics, experimental results show that the proposed method has subjectively significant targets, clear backgrounds and better visual effects in comparison with other fusion methods. Under the average gradient (AG), information entropy (IE), spatial frequency (SF), and mutual information (MI) of the four objective evaluation indicators, on the premise that MI is relatively good, the other three indicators are in the best position,especially for uniformly illuminated camp images, AG and SF are improved by 6. 9% and 4. 8% in comparison with the best values, thus validating the proposed method effectiveness.