Skew estimation for document images

This thesis represents the authora??s efforts in developing a series of skew estimation models for document images, which include the fiducial-line based which relies on the existence of text lines, the convex-hull based which relies on the existence of paragraphs or columns, and the straight-edge based which relies on the existence of non-textual components that have straight edges or lines. These models focus on solving some of the major challenges that any skew estimators still have to face today a?? excessive noises in documents, multiple skews with locations, and scanning artifacts such as edges warping and components touching, and so on. The performance of the three models are evaluated using the full set of samples from the University of Washington English Document Image Database I (UW-I). Quantitative and qualitative comparisons are also provided with the works from University of Washington, Xerox PARC, Siemens AG, AT&T Bell Labs, and Hitachi Central Research Lab.

Author dc:creator

• YUAN BO

• document image analysis; skew estimation; fiducial lines; convex hulls; Muff transform; UW-I