保持結構的影像映射

Abstract

Image remapping has received considerable attention in image processing and computer graphics in recent years. Previous approaches only focus on shape preservation while ignore the area change and important line structures in the image. To achieve structure preservation, image remapping must minimize distortion on not only shapes but also areas and line structures. In this dissertation, we propose four structure-preserving approaches for image remapping. The first one is a content-aware image resizing method which simultaneously preserves both salient image features and important line structure properties: parallelism, collinearity and orientation. The second one is a novel projection model for wide angle image visualization. The proposed model is a mapping from a hemisphere to a plane. Our proposed projection have the following properties: shape preserving, line preserving and rectangular outline. The third one is a projection model for panorama viewing. The proposed projection aims at making a good compromise among the preservation of shapes, area and lines. Both the second the third one are instances of the two-step projection model. The last one is a shape-preserving warp for image stitching. The proposed warp is a spatial combination of a projective transformation and a similarity transformation. such combination simultaneously takes into account the accuracy of alignment and the preservation of image structures.