A video stabilization algorithm is presented based on the feature window tracking. First, the proposed algorithm extracts the corner-points as feature from the specified reference frame. Then, based on the fuzzy Kalman filtering, the extracted features are tracked with window-matching method from the succeed frames. By registering the corresponding feature windows, the motion parameters of the camera can be calculated between the reference frame and succeed frames. Finally, vibration of images due to camera shake is stabilized with motion compensation. The experimental results illustrate that the proposed algorithm is effective and robust with low computational complexity.