A scene-based nonuniformity correction algorithm for scanning-type infrared camera

A scene-based nonuniformity correction algorithm for Infrared line scanner (IRLS) combining constant-statistics approach with neural networks is proposed, correcting the aggregate nonuniformity by two steps. First, the recursive channel-correction based on local constant statistics constraint is performed frame by frame. Secondly, a linear neural network added some optimization strategies is adopted to the result obtained from the first step, making the correction parameters of line sensors update column by column and generating final corrected result at one frame. Applications to both simulated and real infrared data show that the adaptive algorithm has advantages of low computational load and can achieve a higher correction level in less than 20 frames, removing striping noise effectively.