Table of Contents
Authors: Syed Saqib Bukhari, Faisal Shafait, and Thomas M. Breuel
Keywords:
- ridge
- printed text
- non-text segmentation
- gaussian-filter bank
- reading order
Q1: How is line skew determined?
There is a $\theta$ parameter in the Gaussian kernel which is used to produce ridges. This may be used in detecting the skew, but since it’s constant for an entire page, a varying line skew will probably decrease its performance.
Q2: How are non-text portions detected?
The paper does not include a description but cites “S. S. Bukhari, F. Shafait, and T. M. Breuel, ‘Improved document image segmentation algorithm using multiresolution morphology,’ in Proc. SPIE Document Recognition and Retrieval XVIII, San Jose, CA, USA, Jan. 2011” as a source for an improved technique.
Q3: Which heuristics are used in reading order determination?
Breuel is reported to have an algorithm in “T. M. Breuel, ‘High performance document layout analysis,’ in Symposium on Document Image Understanding Technology, Greenbelt, MD, USA, April 2003.” The paper says the authors modified the algorithm for right-to-left scripts. No further details are provided.
Q4: How large is the dataset, and what does it contain?
25 Arabic documents and 20 Urdu documents are used.
Q5: Are there any techniques applicable to divans?
There might be, if any of them were described in detail. We already have more sophisticated text line detection techniques. For the others, I’ll need to read the cited works.