Knowledge-Based Multi-Spectral Pixel Level Fusion for Surveillance

Multi-spectral fusion focused on the task of image enhancement by processing raw data collected at various electromagnetic bands via passive sensors. Passive spectral sensors collect information about the scene based on the inherently reflected or emitted energy of the scene and is represented by spectral distributions and intensities. These systems can be designed so that they provide collocated information through common optics or other means (algorithms). This paper addresses the issue of multi-spectral pixel-level fusion to enhance the visual quality of the "combined" (fused) image data as compared to that of each single spectral band image data. We propose a knowledge-based fusion methodology. We also make the implicit assumption that visually superior image data are also better image data for processing by most types of algorithms.