A Kalman filter in motion analysis from stereo image sequences

In this paper, a Kalman filter-based algorithm for 3-D motion estimation from a stereo image sequence using the unified temporal-spatial optical flow field (UOFF) has been proposed. The modeling problem is discussed first. More consideration has been given to determining the covariance matrices for system noise and sensor noise than the previous works using Kalman filtering in image sequence processing. The newly visible image areas, i.e., the disocclusion issue, which have not been considered in the most of previous works, are handled in our algorithm by using a threshold method. Two experiments are presented to demonstrate the effectiveness of our algorithm