Multiband longwave infrared reflectance removal using blackbody channel prior

Drawing from techniques used to dehaze visible, three-channel RGB images, we propose an approach to separate emissive and reflected radiance in images at short distances with a microbolometer-based longwave infrared (LWIR) camera system. The best case for optimal contrast and with the most descriptive information about the scene in an LWIR image would be where no external sources are radiating toward the scene. We introduce the concept of a blackbody channel prior (BCP) to multiband LWIR imaging to describe pixels that represent objects that behave most similarly to perfect blackbodies with an emissivity near unity. Most LWIR images of outdoor scenes are degraded largely in part by reflected sky path radiance. We can estimate scene radiance with a minimized reflective component producing images with enhanced contrast. Experiments on a number of multiband images are present to demonstrate this spectral-based BCP technique and show its potential for preserving scene information while achieving contrast enhancement.