Far/near infrared adapted pyramid-based fusion for automotive night vision

We discuss the fusion of far infrared (FIR) and near infrared (NIR) image data for an improved automotive night vision device. The two wavebands have differing characteristics when using cameras sensitive to these wavelengths in the automotive environment. It is the aim of our work to fuse the information from these two sensor types to optimize the benefits of both wavebands. The choice of NIR radiation for an automotive night vision system with active illumination has been determined by one automotive manufacturer as the most suitable waveband for a single sensor night vision device. FIR radiation was chosen as a second waveband for fusion because sensors operating at these wavelengths detect passive radiation from objects at temperatures of approximately 300 K. Therefore, in principle, it should be possible to image to the horizon on a dry and humidity free night, and to provide clear images of hot bodies such as pedestrians and vehicles in use. We have chosen pyramid based techniques to fuse the two image types as they allow maximum flexibility, are relatively simple, and computationally efficient. We have adapted pyramid methods to suit our requirements. We demonstrate improved fusion for our particular application