Motion detection/estimation plays a crucial role in dynamic visual tracking. Whether a dynamic visual tracking system can successfully track a moving target closely depends on the quality of motion detection/estimation results. In dynamic visual tracking, the camera used to capture images is not stationary, so any slight vibration of the camera motion or the target motion can lead to a blurry image causing the visual tracking performance to be deteriorated. To cope with this difficulty, a motion detection/estimation approach consisting of a region-based spatial distribution of Gaussians (SDG)-like matching algorithm, a template-update-with-memory algorithm, and a template mask is developed in this study. Moreover, linear interpolation on vision commands is performed to improve the tracking performance. A dynamic visual tracking system designed for locking the target’s image in the center of the image plane is used as the test platform. Experimental results demonstrate the effectiveness of the proposed approach.