Revisiting Far/Near Infrared Pyramid-Based Fusion Types for Night Vision Using Matlab

Abstract

The night vision imaging mechanisms are developed to increase the visibility beyond normal human perception capabilities. So far, night vision methods reported in literature, such as, Morphological, Low Pass Pyramid, Contrast Pyramid, Filter Subtract Decimate and Shift Invariant methods, the Laplacian fusion method has been rated, the best method [1][2]. In this research paper four different methods of fusion of images, Gradient, Wavelet, Quincuns Lifting, including Laplacian are processed using Matlab toolbox called Matifus for night vision. For comparing the results of processed images using above methods, Mean Opinion Score (MOS) is used. MOS result of Laplacian, wavelet, gradient and quincunx methods are compared. The MOS results on the scale of 1-5 indicate a score of 4.15 for Laplacian, that means the quality of image perceived by the scorers is rated between good and excellent. By using MOS and perceptually proving that Laplacian technique is better than all others for night vision systems. However Gradient scored 3.56, Wavelet scored 3.15 and lastly 2.22 was scored by Quincunx Lifting method.