This paper proposes a method for fast panoramic image generation using a rapidly moving robotic vision system. The proposed method consists of 1) real-time image deblurring within 30 frames/s, 2) dynamics-based homography estimation, and 3) parallel system architecture for visual-motor coordination and accelerating computation time. In this study, multiple images were obtained in the course of rapid motion, while a robotic vision system swept an environment. The vision system received blurry images that were recovered by real-time image deblurring. The resultant images were stitched together to generate a panorama image by the dynamics-based approach. The effectiveness of the proposed method was investigated under various motion conditions using a single-degree-of-freedom camera orientation system through quantitative and qualitative comparisons with conventional methods. Experiment results show that real-time performance within 30 frames/s was achieved for both image deblurring and image stitching for VGA-size images. The quality metric of the proposed method was 41% better than conventional methods.