A THINNING-BASED JUNCTION DETECTION AND RESOLUTION ALGORITHM FOR DOCUMENT IMAGES
Junction detection plays a significant role in document image recognition. High recognition rate of
graphical primitives is correlated with the proper detection of junctions. In this paper, a junction detection
algorithm is presented in thinning-based raster to vector conversion process. The method has three stages
that leverage junction representation from pixels to features (i.e. junctions). The input image is thinned to
its skeleton. Edges were found next and pixels with many neighbours are designated as a low level
junction. Polygonal approximation on edges is used to detect L-junctions while connected component
analysis is used to find intermediate junctions. Intermediate junctions with a distance less than a threshold
are combined to form a high level X- and Y-junctions. Performance evaluation on mechanical engineering
drawings shows precision rate of 82.38% and a recall rate of 97.29%.
Three-stage Junction Detection in Document
Images, 2020 International Conference on
Computer Science and Software Engineering
(CSASE), Duhok, Iraq, pp. 142-145, doi:
Di Stefano, L. and Bulgarelli, A. (1999). A simple and
efficient connected components labeling
algorithm. In Proceedings of the International
Conference on Image Analysis and Processing,
pages 322–327, Venice, Italy. Doi:
Hilaire, X. and Tombre, K. (2006). Robust and
accurate vectorization of line drawings. IEEE
Transactions on Pattern Analysis and Machine
Lim, K. B., Xin, K., and Hong, G. S. (1995).
Detection and estimation of circular-arc
segments. Pattern Recognition Letters,
Ma, J., Wang, X., He, Y., Mei, X., and Zhao, J. (2019).
Line-based stereo slam by junction matching
and vanishing point alignment. IEEE Access,
Nakagawa, Y. and Rosenfeld, A. (1979), A note on
polygonal and eliptical approximation of
mechanical parts, Pattern Recognition, 11,
Pham, T.-A., Delalandre, M., Barrat, S., and Ramel, J.
(2012). Accurate junction detection and
reconstruction in line-drawing images. In
Proceedings of the 21st International
Conference on Pattern Recognition
(ICPR2012), pages 693–696, Tsukuba, Japan.
Pham, T.-A., Delalandre, M., Barrat, S., and Ramel, J. (2013). Robust symbol localization based on
junction features and efficient geometry
consistency checking. In 2013 12th
International Conference on Document
Analysis and Recognition, pages 1083–1087,
Washington, DC, USA.
Pham, T.-A., Delalandre, M., Barrat, S., and Ramel,
J.-Y. (2014). Accurate junction detection and
characterization in line-drawing images.
Pattern Recognition, 47(1):282–295.
Seo, W., Koo, H., and Cho, N. (2014). Junction-based
table detection in camera-captured document
images. International Journal on Document
Analysis and Recognition (IJDAR), 18. pages
Wang, Y., Huang, Y., and Huang, W. (2019). Crack
junction detection in pavement image using
correlation structure analysis and iterative
tensor voting. IEEE Access, 7:138094–
Wenyin, L. (2006). The third report of the arc
segmentation contest. In Lecture Notes in
Computer Science, volume 3926 NCS of
Lecture Notes in Computer Science (including
subseries Lecture Notes in Artificial
Intelligence and Lecture Notes in
Bioinformatics), pages 358–361, Hong Kong,
China. Springer Verlag, Heidelberg.
Yuan, J., Chen, H., and Cao, H. (2015). An efficient
junction detection approach for
mobile-captured golf scorecard images.
Procedia Computer Science. 3rd International
Conference on Information Technology and
Quantitative Management, ITQM 2015,
55:792–801, Rio De Janeiro, Brazil.
Zhang, W., Sun, C., Breckon, T., and Alshammari, N.
(2019). Discrete curvature representations for
noise robust image corner detection. IEEE
Transactions on Image Processing,
Zhang, Y. Y. and Wang, P. S. P. (1996). A parallel
thinning algorithm with two-subiteration that
generates one-pixel-wide skeletons. In
Proceedings of the International Conference
on Pattern Recognition, volume 4, pages 457–
461, Vienna, Austria.
It is the policy of the Journal of Duhok University to own the copyright of the technical contributions. It publishes and facilitates the appropriate re-utilize of the published materials by others. Photocopying is permitted with credit and referring to the source for individuals use.
Copyright © 2017. All Rights Reserved.