Combining Bilateral Filtering and Fusion of Visual and IR Images

Abstract

This paper presents efficient fusion algorithms to discover hidden objects in the visual scene by means of merging visual and infrared images of the same scene. Generally, images are corrupted by noise. It is highly difficult to discover the objects in the corrupted image due to different types of noise appearing in the images. To remove noise while preserving edges in noisy input images, bilateral filter is proposed in this paper. Most popular fusion techniques including average and condition rule is employed to obtain a complement fused image from the noisy source images. Along with these two fusion rules, four algorithms have been generated with bilateral filter for finding hidden objects. First, the visual and IR sources degraded by noise are smoothed by bilateral filter. Second, both IR and visual-denoised images are fused by applying one of the proposed pixel-level techniques. The proposed algorithms are tested over four sets of visual and IR images to find out objects that are having worse background as smoke, illumination and bad weather climate. Experiments have been carried out and results were obtained.

Keywords: Image Fusion, Bilateral Filter, Object Detection, Hidden Objects, IR Image, Multisensor

View PDF