Please use this identifier to cite or link to this item:
https://ptsldigital.ukm.my/jspui/handle/123456789/476150Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Siti Norul Huda Sheikh Abdullah, Prof. Dr. | |
| dc.contributor.author | Hayder Ayad Dawood (P56115) | |
| dc.date.accessioned | 2023-10-06T09:14:05Z | - |
| dc.date.available | 2023-10-06T09:14:05Z | - |
| dc.date.issued | 2012-06-01 | |
| dc.identifier.other | ukmvital:74624 | |
| dc.identifier.uri | https://ptsldigital.ukm.my/jspui/handle/123456789/476150 | - |
| dc.description | In arriving new fast technology and image investigation in many applications has made managing visual object categorization (VOC) application extremely important. The main problem of VOC is the semantic gap (categorization problem) which is the gap between the low-level feature and high-level concept in the human mind. However, there are only several ways proposed to overcome this problem and improve its performance. Currently, several researches in the pattern recognition field showed that combining different output classifiers using naive approach and filter banks such as Gabor filter can help in improving the categorization problem. Based on these researches, Gabor filter provides meaningful information and produces distinctive features. Conversely, this research has found that this filter produces incompact and redundant filters which decrease the system performance. Therefore, this research proposes generalization to Gabor filter using unsupervised machine learning algorithm denoted by K-means technique to produce optimize and compact filters and remove the redundant filters. This research also focuses on combining different filters using naive approach with and without the proposed method. The edge histogram of MPEG-7 descriptor is used to extract the texture feature and the SVM algorithm is applied for classification purpose. Furthermore, the VOC technique based on the proposed method and the standard Gabor filter has been performed in the spatial and frequency domains to explore its performance. The first 20 categories and the whole categories of Caltech 101 dataset are proposed to test and evaluate the performance of the proposed method. Based on a single classifier and combination feature (naive approach), the proposed method outperforms and shows higher potential results than the standard method in describing objects.,Master/Sarjana | |
| 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 | Gabor filter | |
| dc.subject | Visual object categorization | |
| dc.subject | Computer vision -- Mathematics | |
| dc.title | Generalization of gabor filter for visual object categorization | |
| dc.type | theses | |
| dc.format.pages | 122 | |
| dc.identifier.callno | TA1637.D346 2012 3 | |
| dc.identifier.barcode | 000452 | |
| Appears in Collections: | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| ukmvital_74624+Source01+Source010.PDF Restricted Access | 2.4 MB | Adobe PDF |  View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.