Non-Uniformity Correction Algorithm Suite (NUCAS)

Warfighters depend on infrared focal plane array (IRFPA) sensors to detect and recognize targets in many situations. Unfortunately, modern IRFPAs still suffer from sensor non-uniformity, which can be even more limiting than low contrast or lack of resolution. To make matters worse, the non-uniform response exhibits temporal drift, making it impossible to perform a one-time calibration at the factory. There are two common techniques for non-uniformity correction: radiometric (which requires that the sensor be taken off-line for calibration) and scene based (using just the sensor data). Current scene-based non-uniformity correction techniques have limitations for static scenes as well as for dynamic scenes exhibiting complex motion. We propose to develop a suite of algorithms for non-uniformity correction. The starting point for this effort will be our real-time autonomous video enhancement system, which is capable of correcting numerous single- and multi-frame artifacts. We will use this system to develop algorithms for motion-based non-uniformity correction and for scene-based correction that is also radiometrically accurate. We employ a motion estimation algorithm that is sub-pixel accurate, can handle arbitrary changes in the scene and is robust to large amounts of noise. Lastly, we develop methods for handling scenes without any apparent motion.