Nonsubsampled Contourlet Domain Fusion Approach for Infrared and Visible Fire Images

Abstract

The recurring forest fires have prolonged catastrophic effects on the environment as well as human individuals. The detection of such fire is a very prominent issue using visible and infrared (IR) images. This paper presents a fusion approach for IR and visible images using several features of the nonsubsampled contourlet transform (NSCT). Firstly, NSCT is applied to decompose the reference IR and visible image into the different low and high-frequency components. Low-frequency coefficient is fused by a pulse coupled neural network (PCNN) motivated by the sum-modified Laplacian (SML) to retain the more amount of information available in both the source images and log Gabor energy based fusion rule is applied to fuse the high-frequency coefficients to preserve the more edge details. At last, inverse NSCT is applied to reconstruct the fused image. Furthermore, several experiments are performed to evaluate the performance of the proposed approach in terms of visual appearance as well as several performance measures. Experimental results show the superiority of the presented fusion method in the NSCT domain over the other existing fusion approaches in terms of the improvement in all the performance measures.