Please use this identifier to cite or link to this item:
https://ptsldigital.ukm.my/jspui/handle/123456789/513282Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Khairuddin Omar, Prof. Dr. | - |
| dc.contributor.author | Jabril Ramdan Abdslam Salem (P63460) | - |
| dc.date.accessioned | 2023-10-16T04:35:14Z | - |
| dc.date.available | 2023-10-16T04:35:14Z | - |
| dc.date.issued | 2017-03-06 | - |
| dc.identifier.other | ukmvital:97590 | - |
| dc.identifier.uri | https://ptsldigital.ukm.my/jspui/handle/123456789/513282 | - |
| dc.description | In the Arabic Character Recognition (ACR) process, the segmentation is regarded as being among the most difficult and tiresome processes. The most intricate part is within character segmentation because it entails bulges or strokes that are simple to segment, while others are not, such as س and ص. In addition, characters such as ل and ك are equally hard to segment due to the fact that they entail a tail segment. The present ACR approaches are characterized by simplicity and, essentially, are short of the mathematical foundation. They seem to disregard the solid basis of illustrating characters and their associated points, unable to reflect on Arabic characters’ particular individual characteristics and they employ inadequate attributes because of their use of pixel-based approaches. Several techniques and approaches to segmentation exist, for instance, segmentation free and segmentation based. This concern can be handled and categorized under the segmentation based approach. Presently, the frequently utilized approach has been Voronoi Diagrams (VD). Divide and Conquer (DAC) and VD signify mathematical background and regions and are able to handle segment-linked constituents based on neighbours and elevate to predict Vertexes (VX) points to sub-words and segment words. There exist key and valuable notions in computational geometry and in a wide range of applications. This study intends primarily to suggest an algorithm to segmentation model using neighbourhood information which consists of four sub objectives as: determine neighbour graph of image’s connected components, propose a mathematical approach that defines a region-based segmentation approach for ACR, propose an algorithm to determine segmentation points depending on minimum peak with baseline and reconfirm its correct location points with Vertexes points, and establish and enhance a segmentation approach based on tracing of VX and peaks in order to isolate sub-words or Arabic words to characters. Secondly, it produces and collects a new Arabic handwritten dataset with ground truth to help and improve recognition stage. Third it compares the proposed technique with the state-of-art. The construction of neighbour graphs of associated components is founded on VD algorithm, a line is drawn from the middle of the connected components to mark out the lower, centre, and the upper locations of the neighbours, the white background, hence regarded as a fully neighbour graph. If a point of connected component boundary traverses amid the partly neighbours, or if the connected component realized amid the other two boundaries is not a white background, then they cannot be termed as neighbours. The Euclidean distance is utilized as a base to sketch the edge segment amid the linked components, hence coming up with area VD. The following stages will be covered in the proposed region-based segmentation: i) Computing for minimum and maximum peaks; ii) Execution of direction method, and iii) Enhancement of the model utilizing the projected mathematical model. For the purposes of enhancing segmentation, selection of segmentation points through minimum peaks will be accomplished by use of edges and VX points. This study equally aims at advancing the current research on the assessment of Arabic and Malay manuscripts. Empirically, this approach estimates to promising results with a high accuracy to segmentation Arabic characters via the use of AHDB, IFN-ENIT, AHDB-FTR, APTI, and ACDAR DB datasets. The primary results, it is evident that the use of MNN and EDMS facilitates increased valuable findings than MOMENT and GLCM in further classifications producing results of 98.27% for the APTI printed dataset and 95.09% for the IFN-ENIT dataset.,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 | Arabic characters | - |
| dc.subject | Voronoi polygons | - |
| dc.title | Segmentation methods of Arabic handwriting using neighbourhood information for voronoi diagrams | - |
| dc.type | Theses | - |
| dc.format.pages | 216 | - |
| dc.identifier.callno | QA278.2.S438 2017 3 tesis | - |
| dc.identifier.barcode | 002958(2017) | - |
| Appears in Collections: | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| ukmvital_97590+SOURCE1+SOURCE1.0.PDF Restricted Access | 743.6 kB | Adobe PDF |  View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.