| Author | Baber, Junaid |
| Call Number | AIT Diss. no.CS-13-07 |
| Subject(s) | Web search engines
|
| Note | A dissertation submitted in fulfilment of the requirements for the degree of Doctor of Philosophy in Computer Science, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Abstract | Test based search engines are getting famous, fast, and accurate despite the exponential growth in data. There are many search engines like Google, Bing and Yahoo which provides interactive and fast interface for searching; these searches are mainly based on text based key words and phrases. Now-a-days data and information is transformed in multimedia format such as images and videos. Multimedia contents are used for education, information and entertainment. These data particularly images are also increasing exponentially over Internet. People (users) are interested to query the search engine by some visual cues and retrieve the images and videos based on those cues, on the other hand, companies are interested to prevent the illegal distribution of forged and copyrighted images. There is much work reported for retrieval of images specially image copies, but still searching the query image or object is computationally very expensive, inaccurate and impractical like text based searches. The most famous criteria for image copy retrieval is to represent the images by set of local key point descriptors which should be robust, efficient, and discriminative. The problem to extract these descriptors is the main challenge. There are many descriptors but all descriptors lack either in efficiency, robustness or distinctiveness. Our framework increases the performance, particularly robustness and distinctiveness, of local key point descriptors for large scale image copy retrieval and video segmentation. We also propose binary quantization of gradient histograms which is very simple but effective. This quantization can be applied for image copy retrieval for small scale and resource limited devices such as smart phones and tablets. This thesis describes the state-of-the art for image copy retrieval and video segmentation; we also present effective techniques to improve the performance on challenging and benchmark datasets. |
| Year | 2013 |
| Type | Dissertation |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Afzulpurkar, Nitin; |
| Examination Committee(s) | Martinet, Philippe ;Dailey, Matthew N. ;Guha, Sumanta; |
| Scholarship Donor(s) | University of Balochistan (UOB), Quetta, Pakistan;Asian Institute of Technology Fellowship; |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2013 |