Scene-based nonuniformity correction algorithm based on optical flow

Abstract

The performance of infrared imaging system is strongly affected by non-uniformity in infrared focal-plane arrays (FPA). In the classical scene-based nonuniformity correction (NUC) method, errors commonly occur resulting from local motion between two frames. In this paper, a novel scene-based NUC method is presented. This method calculates robust optical flow between two adjacent frames to get the velocity vector of each pixel in the current frame. In this way, corresponding to the pixel in the current frame, the location of the pixel in previous frame is known, and then these frames can be locally registered easily. Based on the assumption that any two detectors with the same scene would produce the same output value, minimize the mean square error between two local registered images to get the estimation of each detector’s gain and offset. With gain and offset parameters, nonuniformity of infrared imaging system can be corrected. One advantage of this scene-based NUC algorithm is that it can adapt to scene with local motion. The performance of the proposed algorithm is studied with infrared image sequences with simulated nonuniformity and infrared imagery with real nonuniformity. It shows that fixed-pattern noise is reduced efficiently even when the scene include local motion.