A general edge detection procedure is combined with new uses of temporal information in a segmentation algorithm for time-varying imagery. Edge detection is performed by thinning and thresholding the Prewitt gradient. Gaps in object boundaries are then filled by temporal information obtained by a multi-frame process of filtering, edge enhancement, and edge linking. The performance of the algorithm is tested by application to both FLIR and synthetic data. The FLIR data was collected by a sensor on a moving aerial platform recording a target that was also in motion. The synthetic data was created to model this type of complex motion in a controlled environment. The success of the method in segmenting target data and its advantages over conventional single frame techniques is discussed and illustrated.