Please use this identifier to cite or link to this item:
https://ptsldigital.ukm.my/jspui/handle/123456789/513283Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Khairuddin Omar, Prof. Dr. | |
| dc.contributor.author | Waleed Abdel Karim Abu-Ain (P49057) | |
| dc.date.accessioned | 2023-10-16T04:35:14Z | - |
| dc.date.available | 2023-10-16T04:35:14Z | - |
| dc.date.issued | 2016-01-05 | |
| dc.identifier.other | ukmvital:97638 | |
| dc.identifier.uri | https://ptsldigital.ukm.my/jspui/handle/123456789/513283 | - |
| dc.description | Skeleton extraction is an important preparation process for many pattern recognition applications. However, simple sequential thinning methods available at present are still suffering of many thinning challenge problems. They fail in extracting one-pixel width skeleton, preserve topology, preserve connectivity and preventing the spurios tails in all rotation angles. Regardless of the above, feature extraction is another fundamental process in pattern recognition. Usually, statistical analysis techniques shortcoming occur when dealing with binary document images due to less discriminative values are available in both first and second order approaches. Thus, a higher order statistics is required to find additional properties. Automatic Optical Script Recognition (AOSR) is one of the main challenges in document image analysis and recognition domain. Currently, only a few attempts were stated for automated script identification of off-line handwritten documents images. Most available AOSR applications only deal with printed documents and script types, and they neglect handwritten and multi-lingual documents. In this work, three objectives are highlighted, first is to propose a thinning method for preprocessing image, second is to propose a statistical texture analysis technique via binary images for feature extraction and third is to propose a multi-lingual AOSR framework. The research methodology consists of three proposed ideas namely a proposed generic thinning algorithm for skeletonisation method based on sequential iterative approach, a proposed statistical texture analysis method for binary images via Skeleton Primitive Direction Matrixes (SPDM), and a proposed multilingual AOSR framework. The proposed thinning method has been tested on DIBCO 2010, MPEG-7 CE-Shape-1 benchmark and self-collected datasets. Its performance has also been evaluated and compared with state of the art methods namely Zhang-Suen, Huang, and K3M methods. Based on experimental results of thinning measurement, the proposed thinning method gained 99.69%, 99.99% and 99.90% whereas, Zhang-Suen method achieved 92.63%, 99.40% and 97.3%, Huang method achieved 98.95%, 99.95% and 99.60 and K3M method achieved 85.80%, 97.71% and 91.50% for each dataset respectively. Next, SPDM method is tested on Arabic calligraphy and EngFont benchmark datasets. Its performance is also compared with state of the art methods namely Gray-Level Co-occurrence Matrix (GLCM), Local Binary Pattern (LBP), Edge Direction Matrixes (EDMS) and Gabor Filter. From experimental point of view, the accuracy of SPDM outperformed GLCM, LBP, EDMS and Gabor in both datasets 97.51%, 89.49%, 84.8%, 93.67% and 97.2% respectively for Arabic calligraphy dataset, and 90.6%, 45.14%, 86.85%, 46.5% and 79.90% respectively for EngFont dataset. Finally, the proposed multilingual AOSR framework tested on Multilingual-HW datasets, which contain more than seven international unconstraint handwritten scripts, and the overall average accuracy is about 98.87%.,Certification of Master's/Doctoral Thesis" is not available | |
| dc.language.iso | eng | |
| dc.publisher | UKM, Bangi | |
| dc.relation | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat | |
| dc.rights | UKM | |
| dc.subject | Skeleton extraction | |
| dc.subject | Optical pattern recognition | |
| dc.title | Automatic optical script recognition based on sequential iterative skeletonisation and primitive direction matrix features | |
| dc.type | Theses | |
| dc.format.pages | 217 | |
| dc.identifier.callno | TA1650.A256 2016 3 tesis | |
| dc.identifier.barcode | 002971(2017) | |
| Appears in Collections: | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| ukmvital_97638+SOURCE1+SOURCE1.0.PDF Restricted Access | 3.18 MB | Adobe PDF |  View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.