Finding Corresponding Patches in Texture Images Using Tensor Consistency Check

摘要

In this paper, we propose a method for finding the corresponding patches across multiple images using tensor consistency check. We assume image correspondences are unknown and the camera sensitivity parameters on intensity are uncalibrated. A patch is defined to be a 2D array/grid of points on a 3D surface. We assume a set of patches are obtained for each view, but the correspondence of individual patches across multiple images is not known. For finding correspondence of these patches, we consider a camera projection model which combines the geometric and photometric information of objects. Furthermore, we consider camera projections in frequency space. By deriving multi view relations in frequency space, we show that we can correctly find the corresponding image patches of 3D surfaces across multiple images. The efficiency of the proposed method is tested using real images.