Infrared and visible image fusion using latent low rank technique for surveillance applications

Abstract

Image fusion aims at the integration of different complementary image data into a distinct, new image with the best achievable quality. Fusion of visible and infrared images provides complementary performance which is frequently required in many standard vision-based systems. For example, military and surveillance systems require target detection (thermal) followed by identification (visible); Comparative analysis of different fusion techniques with Latent low rank method (LLR) is done on different military and surveillance applications. In case of concealed weapon detection, LLR performance is good, where as DWT based fusion techniques are suitable for surveillance applications but in case of certain data sets feature extraction is not appropriate. In this paper, Latent low rank method, which is an accurate technique for Image fusion to find hidden weapons or other objects hidden beneath an individual’s clothing, is presented. LLR technique is implemented using MATLAB-2019 tool. Latent low rank representation has the power to spot salient features. This particular model de-noises and decomposes the image simultaneously. This method is simple and effective. The percentage of detection of objects is 94.6%. Different metrics are used for evaluating fusion performance subjectively. Simulation results and subjective evaluation shows that LLR is more suitable for concealed weapon detection application.