Parallax estimation for push-frame satellite imagery: application to super-resolution and 3D surface modeling from Skysat products

Recent constellations of satellites, including the Skysat constellation, are\nable to acquire bursts of images. This new acquisition mode allows for modern\nimage restoration techniques, including multi-frame super-resolution. As the\nsatellite moves during the acquisition of the burst, elevation changes in the\nscene translate into noticeable parallax. This parallax hinders the results of\nthe restoration. To cope with this issue, we propose a novel parallax\nestimation method. The method is composed of a linear Plane+Parallax\ndecomposition of the apparent motion and a multi-frame optical flow algorithm\nthat exploits all frames simultaneously. Using SkySat L1A images, we show that\nthe estimated per-pixel displacements are important for applying multi-frame\nsuper-resolution on scenes containing elevation changes and that can also be\nused to estimate a coarse 3D surface model.\n