Paper stitching using maximum tolerant seam under local distortions

Abstract

Paper stitching technology can reconstruct a whole paper page from two sub-images separately scanned from a camera with limited vision field. Traditional technology usually chooses a global optimal seam, and the two sub-images are stitched along it. These methods perform well on the rigid object, but when distortion exists caused by the uneven placement of paper, local contents of two sub-images may be upside-down and their positions are misaligned. Although some methods choose two matching seams on each sub-image, they use either the local patch similarity or the global consistent constraint to get two matching seams. However, only the local matching may lead to stitching failure when wrong matching occurs at the local patch, while only the global constraint usually suffers from inaccuracy of the stitching result. After the two seams are obtained, the traditional methods usually construct the whole image through global transformation along the seams, and image deformation usually occurs in this stage. In this paper, we proposed a robust estimation algorithm to get the matched seams in the sub-images, and stitched the sub-images with a maximum tolerance to conquer the image deformation. Finally a whole image with a smooth stitching seam and the minimum deformation is generated. Experimental results show that this new paper stitching method can produce better results than state-of-arts methods even under challenging scenarios such as large distortion and large contrast difference.