Photo zoom: high resolution from unordered image collections
Martin Eisemann, Elmar Eisemann, Hans‐Peter Seidel, Marcus A. Magnor · 2010
Figure 1: Our technique allows infinite zooms into images, creating new details at each level. Not only from single images, but also from image collections. From left to right: zoom into an image with original resolution of 256 × 256 using Photo Zoom, bilinear upsampling and nearest neighbor upsampling of the same region. We present a system to automatically construct high resolution im-ages from an unordered set of low resolution photos. It consists of an automatic preprocessing step to establish correspondences be-tween any given photos. The user may then choose one image and the algorithm automatically creates a higher resolution result, sev-eral octaves larger up to the desired resolution. Our recursive cre-ation scheme allows to transfer specific details at subpixel positions of the original image. It adds plausible details to regions not cov-ered by any of the input images and eases the acquisition for large scale panoramas spanning different resolution levels.