This paper describes a new versatile algorithm for cor- recting nonlinear distortions, such as curvature of book pages, in camera based document processing. We introduce the idea of using local orientation features to interpolate a vector field from which a warping mesh is derived. Ulti- mately, the image is corrected by approximating the non- linear distortion with multiple linear projections. Since the algorithm does not derive the mesh directly from text base- lines it is robust over arbitrarily complex text layouts. We describe a baseline detector for extracting the required lo- cal orientation features. We also sketch a method for cor- recting nonlinear distortions of a document's vertical axis with our algorithm.
Citation:
D. Schneider, M. Block, R. Rojas, "Robust Document Warping with Interpolated Vector Fields," icdar, vol. 1, pp.113-117, Ninth International Conference on Document Analysis and Recognition (ICDAR 2007) Vol 1, 2007