FrameBreak: Dramatic Image Extrapolation by Guided Shift-Maps

Abstract

In this thesis, we propose a method to significantly extrapolate the field of view of a photograph by learning from a roughly aligned, wide-angle guide image of the same scene category. Our method can extrapolate typical photos into complete panoramas. The extrapolation problem is formulated in the shift-map image synthesis framework. We analyse the self-similarity of the guide image to generate a set of allowable local transformations and apply them to the input image. Our method could preserves to the scene layout of the guide image when extrapolating a photograph. To expressively characterize the self-similarity of complex scenes, image extrapolation system additionally allows image transformations of rotation, scaling and reflection. To handle this increase in complexity, we introduce a hierarchical graph optimization method to choose the optimal transformation at each output pixel. We demonstrate results on a variety of indoor, outdoor, natural, and man-made scenes.