Preemptive text warping to prevent appearance of motion blur

Zixun Yu, Manuel M. Oliveira, Daniel G. Aliaga

Abstract: We present a preemptive image-based method to reduce motion blurring. Motion blur appears when there is a relative motion between the scene and the viewer/camera. A preemptive method pre-filters the content before being displayed in order to mitigate the occurrence of motion blur. Our experiments and user study have shown that such preemptive methods are fundamentally subject to producing visual artifacts as a consequence of unavoidable ringing or intensity oscillations. Frequency-domain analysis shows that the energy weakened at certain frequencies leads to those artifacts. We present a method to process alphanumeric content so that it has lower energy on frequencies eliminated by a given motion blur kernel. Our processed image, when motion blurred, will have a sharper appearance and less artifacts as compared to various alternative approaches. We demonstrate the effectiveness of the proposed technique with simulated and real-world experiments as well as user feedback. Our results show that our approach yields content robust to motion blur while still being perceptually similar to the original text.