Conceptual Design for Video Rate Pixon Image Processing Hardware

The astronomical community recognizes the Pixon(TM) method as the highest performance image reconstruction method ever developed. In addition to tremendous performance gains relative to competing methods (factors of several in resolution, orders of magnitude in sensitivity and robust artifact rejection), the Pixon method is much faster than standard non-linear methods (orders of magnitude). By nature, the Pixon method is highly parallelizable, suggesting that hardware implementation should achieve video rates (30 Hz) for 512x512 pixel images. We propose to design such hardware, as a prelude to construction in Phase-II. Such hardware would have tremendous commercial applications (medical imaging, radar, commercial satellite and aircraft imaging, surveillance, military intelligence), and be directly relevant to the goals of NASA's Data Understanding and Adaptive Methods (SBIR 99-1, Topic 21.03) program. The Pixon method directly provides the ability to deal quantitatively with uncertainty present in data by building flexible models through which observables are linked to quantities of scientific or engineering interest. A Pixon hardware implementation would be of immediate benefit to NASA, dramatically improving the quality of NASA's imaging data. Furthermore it will provide a natural front-end processor for future robotic vision or image classification/segmentation engines.