Current infra-red focal point arrays are limited by their inability to calibrate out component variations. Nonuniformity correction (NUC) techniques have been developed and implemented in off-board digital hardware to perform the necessary calibration for most IR sensing applications. There are two possible types of NUC that can be considered for focal-plane integration: (1) Two-point correction using calibrated images on startup and (2) Scene- based techniques that continually recalibrate the sensor for parameter drifts. The problems with the two-point methods have been well-documented in the literature (parameter drift, expense, etc.) We address the two major problems of scene-based techniques: (1) a more difficult hardware implementation and (2) ghosting artifacts in the corrected images. We have previously addressed the implementation problems by developing and demonstrating special purpose analog hardware as well as an efficient digital algorithm that incorporates the constant statistics model. The ghosting artifact occurs in all scene-based techniques when an object that does not move enough tends to `burn in' and can remain visible for thousands of images after the object has left the field of view. We have improved our model to eliminate much of the ghosting artifact that plagues all scene-based NUC algorithms. By modifying the correction update during ghosting situations, we are able to significantly remove the ghosting artifact and improve the overall accuracy of the correction procedure. We demonstrate these results on real and synthetic image sequences.
展开▼
