Search

Camera gimbal systems are important in various air or water borne systems for applications such as navigation, target tracking, security and surveillance. A higher steering rate (rotation angle per second) of gimbal is preferable for real-time applications since a given field-of-view (FOV) can be revisited within a short period of time. However, due to relative motion between the gimbal and scene during the exposure time, the captured video frames can suffer from motion blur. Since most of the post-capture applications require blurfree images, motion deblurring in real-time is an important need. Even though there exist blind deblurring methods which aim to retrieve latent images from blurry inputs, they are constrained by very high-dimensional optimization thus incurring large execution times. On the other hand, deep learning methods for motion deblurring, though fast, do not generalize satisfactorily to different domains (e.g., air, water, etc). In this work, we address the problem of real-time motion deblurring in infrared (IR) images captured by a gimbal-based system. We reveal how a priori knowledge of the blur-kernel can be used in conjunction with non-blind deblurring methods to achieve real-time performance. Importantly, our mathematical model can be leveraged to create large-scale datasets with realistic gimbal motion blur. Such datasets which are a rarity can be a valuable asset for contemporary deep learning methods. We show that, in comparison to the state-of-the-art techniques in deblurring, our method is better suited for practical gimbal-based imaging systems.