Super-Resolved Image Synthesis from Uncalibrated Camera with Unknown Motion

In this paper, we propose a novel method for synthesizing super-resolved images from image sequence taken with uncalibrated moving camera by integrating all frames in the sequence based on projective geometry between the frames. In this method, assuming that the object scene can be approximated with multiple plane patches, all frames in the input image sequence are registered with the fundamental matrices between the frames, and then integrated by transforming with homographic matrices. For reducing the registration error, the blur mount in the integrated image is estimated so that the sub-pixel order registration between the frames can be achieved. For demonstrating the effectiveness of the proposed method, we confirm that the proposed method improves the resolution of the synthesized image comparing with the input image sequence.