Application of the progressive wavelet correlation for image recognition and retrieval from the collection of images - Thesis

Stojanovic, Igor (2011) Application of the progressive wavelet correlation for image recognition and retrieval from the collection of images - Thesis. PhD thesis, Faculty of Natural Sciences and Mathematics.

[img]
Preview
Text
PhD_info.pdf

Download (877Kb) | Preview

Abstract

An algorithm for recognition and retrieval of image from image collection is developed. Basis of the algorithm is the progressive wavelet correlation. The final result is the recognition and retrieval of the wanted image, if it is in the image collection. Instructions for the choice of correlation threshold value for obtaining desired results are defined. The areas where the algorithm can be applied are also discussed. To increase efficiency is presented two phases solution. The first phase uses well-known methods of image retrieving by descriptors based on the content of the searched image. In the second phase the progressive wavelet correlation method is applied on the small number of image candidates selected in previous phase. Experiments are performed with data bases of 1000 and 10 000 images, using Oracle data base and the Matlab component Database Toolbox for operations with data bases. The algorithm is applicable to different formats of images. A method for image restoration is presented which is used before the images are being inserted in storage systems. Advantages of the method over other methods are increasing the values of the parameter Improvement in Signal to Noise Ration and reduce the time required to complete the process of restoration.

Item Type: Thesis (PhD)
Subjects: Natural sciences > Computer and information sciences
Engineering and Technology > Electrical engineering, electronic engineering, information engineering
Divisions: Faculty of Computer Science
Depositing User: Igor Stojanovik
Date Deposited: 05 Mar 2014 10:58
Last Modified: 05 Mar 2014 10:58
URI: http://eprints.ugd.edu.mk/id/eprint/9369

Actions (login required)

View Item View Item