Publication
Combined Orientation and Skew Detection Using Geometric Text-Line Modeling
Joost van Beusekom; Faisal Shafait; Thomas Breuel
In: International Journal on Document Analysis and Recognition, Vol. 13, No. 2, Pages 79-92, Springer, 2010.
Abstract
In large scale document digitization, orientation detection plays an important role, especially in
the scenario of digitizing incoming mail. The heavy use
of automatic document feeding (ADF) scanners and
more over automatic processing of facsimiles results in
many documents being scanned in the wrong orientation. These misoriented scans have to be corrected, as
most subsequent processing steps assume the document
to be scanned in the right orientation. Several existing
methods for orientation detection use the fact that in
Latin script text, ascenders are more likely to occur
than descenders. In this paper, we propose a one-step
skew and orientation detection method using a well established geometric text-line model. The advantage of
our method is that it combines accurate skew estimation with robust, resolution independent orientation detection. An interesting aspect of our method is that
it incorporates orientation detection into a previously
published skew detection method allowing to perform
orientation detection, skew estimation, and, if necessary, text-line extraction in one step. The effectiveness
of our orientation detection approach is demonstrated
on the UW-I dataset, and on publicly available test images from OCRopus. Our method achieves an accuracy
of 99% on the UW-I dataset and 100% on test images
from OCRopus.