Robust object tracking using mean shift and fast motion estimation

Visual object tracking is still a challenging problem in computer vision. We use color-based mean shift (MS) tracking algorithm to track object. Meanwhile, we use Kalman to predict the initial location for MS tracker. Predictor can give a reasonable initial position that is close to the target location. This prediction may reduce mean shift iteration number. Then in order to make it more robust, we bring the fast motion estimation (FME) idea used in the video compression field into our system. It acts as a complementary technology to make the tracking result more robust when there exists large object displacement between two adjacent frames.