doi:10.1016/S0262-8856(01)00071-3 
Copyright © 2001 Elsevier Science Ltd. All rights reserved.
Efficient skew estimation and correction algorithm for document images
H. K. Kwag
, 
, S. H. Kim, S. H. Jeong and G. S. Lee
Department of Computer Science, Chonnam National University, PO Box 500-757, 300 Yongbong-Dong, Bug-Gu, Kwangju, South Korea
Received 11 May 2000;
revised 28 March 2001;
accepted 6 June 2001
Available online 4 October 2001.
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Abstract
In this paper, we propose a fast skew estimation and correction algorithm for English and Korean documents based on a BAG (Block Adjacency Graph) representation. BAG is one of the most efficient data structures for extracting various information concerning connected components; the image rotation for skew correction is performed rapidly using the block information in the BAG. The proposed skew estimation algorithm uses a coarse/refine strategy based on the Hough transformation of connected components in the image. The skew correction algorithm then generates a non-skew image by rotating the blocks, rather than the individual pixels. An experiment using 2016 images from various English and Korean documents demonstrates how the proposed method is superior to conventional ones.
Author Keywords: Skew estimation; Skew correction; BAG; Hough transform; Block rotation
Fig. 1. Block diagram of the proposed skew estimation and correction algorithm.
Fig. 2. The creation of a BAG and its bounding box: (a) a connected component; (b) its BAG and bounding box.
Fig. 3. Algorithm for generating a BAG.
Fig. 4. An example result of the data extraction step: (a) original image; (b) bounding boxes; (c) selected bounding boxes.
Fig. 5. An algorithm for the coarse estimation step.
Fig. 6. An example of the coarse skew estimation step.
Fig. 7. An algorithm for the refine estimation step.
Fig. 8. Two adjacent blocks and corner coordinates.
Fig. 9. An example of block filling based on the intersecting point calculation: (a) the intersecting points and inside pixels; (b) the filled block.
Fig. 10. Some of the original skew images.
Fig. 11. Some of the resulting images generated by the proposed algorithm.
Fig. 12. Examples of skew-corrected images: (a) original input images (skewed by +10°); (b) direct method; (c) indirect method; (d) the proposed method.
Fig. 13. The processing time of the three methods for 50 images from English magazines.
Table 1. Number of document images for nine different kinds of skew angles
Table 2. Performance comparison of the skew estimation methods
Abstract
Purchase PDF (797 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (15)
Purchase PDF (442 K)
