DocumentCode
1635450
Title
Registration and Enhancement of Double-Sided Degraded Manuscripts Acquired in Multispectral Modality
Author
Tonazzini, Anna ; Bianco, Gianfranco ; Salerno, Emanuele
Author_Institution
Ist. di Scienza e Tecnol. dell´´Inf., Consiglio Naz. delle Ric., Pisa, Italy
fYear
2009
Firstpage
546
Lastpage
550
Abstract
We propose a system to process multispectral scans of double-sided documents. It can co-register any number of recto and verso channel maps, and reduce the bleed-through/show-through distortions by exploiting blind source separation. From RGB scans, it is also able to recover the original colors, thus improving the readability of a document while maintaining its original appearance. The recto and verso patterns obtained can then be further analyzed. Many approaches to this problem are based on single-channel or multichannel recto-verso scans. In any case, getting rid of the unwanted interferences is a challenging problem. All the methods relying on pixel intensities, such as the one presented here, need a very accurate co-registration, and this is difficult for recto-verso pairs since the relevant information is often very sparse.
Keywords
blind source separation; distortion; document image processing; image colour analysis; image enhancement; image registration; statistical analysis; RGB scan; bleed-through/show-through distortion; blind source separation; document readability; double-sided degraded manuscript; double-sided document; image color analysis; image enhancement; image registration; multichannel recto-verso scan; multispectral modality; recto channel map; single-channel recto-verso scan; statistical analysis; unwanted interference; verso channel map; Blind source separation; Data models; Degradation; Gray-scale; Humans; Image analysis; Image color analysis; Information analysis; Interference; Source separation;
fLanguage
English
Publisher
ieee
Conference_Titel
Document Analysis and Recognition, 2009. ICDAR '09. 10th International Conference on
Conference_Location
Barcelona
ISSN
1520-5363
Print_ISBN
978-1-4244-4500-4
Electronic_ISBN
1520-5363
Type
conf
DOI
10.1109/ICDAR.2009.131
Filename
5277597
Link To Document