Infrared and Visible Image Fusion in Realistic Streetscape

Infrared and visible image fusion plays an important role in military and civilian applications. However, most methods are focus on simple general scene, which only contains one person alone or a few buildings. It is not enough to fully reflect the diversity and complicated of reality. This paper presents a new fusion method for infrared and visible images based on saliency detection and guided filter, aiming to obtain a synthetic image that is suitable for human visual perception in complicated street scenes. The proposed method employs the saliency detection to extract target region from infrared images. Each image is decomposed into a base layer and detail layers based on a multi-scale decomposition through guided filter. Experimental results demonstrate that our method can achieve a superior performance compared with previous image fusion methods in both subjective and objective evaluation.