Key points are not available for this paper at this time.
An important part of any document recognition system is detection of skew in the image of a page. This paper presents a new, accurate and robust skew detection algorithm based on a method for finding rows of text in page images. Results of a comparison of the new algorithm against Baird's well-known algorithm on 400 pages show the new algorithm to be more accurate, robust and somewhat faster. In particular, the new algorithm only breaks down at skew angles in excess of 15 degrees, compared to the almost uniform distribution of breakdowns of Baird's algorithm.
Robert W. Smith (Tue,) studied this question.