Infrared Technology XIX
Results are presented of noise reduction by motion compensated temporal filtering in a noisy IR image sequence and of moving target detection in an air-to-ground IR image sequence. In the case of motion compensated temporal filtering our approach consists of estimating the optical flow between successive frames and subsequently averaging a small number of images. Moving targets are detected by first estimating the optical flow between successive frames. Target detection amounts to comparing a predicted frame, based on the estimated optical flow, to the actual frame. Thus, it is possible to detect targets without making assumptions on their appearance. The particular motion estimator used was found to be especially useful in the case of IR imagery, because the estimator is relatively insensitive to noise and global brightness variations.