Stable Scene-based Non-uniformity Correction Coefficients for Hyperspectral SWIR Sensors

Scene-based non-uniformity correction (NUC) methods commonly produce artifacts as a result of NUC coefficient biasing by the specific scene content (e.g., streaking at high-contrast boundaries). We propose and evaluate a new scene-based method for computing stable non-uniformity correction coefficients for short-wavelength infrared (SWIR) scanning hyperspectral sensors relying on the spatial ratio of spectral ratios to eliminate bias from the image scene. The new technique produces NUC coefficients computed from scene data that converge more quickly and remain more stable than other methods, resulting in calibrated images without scene-induced artifacts. Advantages of our technique are discussed and demonstrated with data from several different airborne sensor systems.