Attention-driven image pre-classification and retrieval

Pao Yue-kong Library Electronic Theses Database

Attention-driven image pre-classification and retrieval

 

Author: Tang, Yu
Title: Attention-driven image pre-classification and retrieval
Degree: M.Sc.
Year: 2009
Subject: Hong Kong Polytechnic University -- Dissertations.
Image processing -- Mathematics.
Digital images -- Classification.
Image transmission.
Department: Dept. of Electronic and Information Engineering
Pages: x, 81 leaves : ill. (some col.) ; 30 cm.
Language: English
InnoPac Record: http://library.polyu.edu.hk/record=b2305940
URI: http://theses.lib.polyu.edu.hk/handle/200/3478
Abstract: This dissertation presents attention-driven image pre-classification and retrieval. Two main contributions are reported in this dissertation, which include: (1) image database pre-classification; and (2) image retrieval with pre-classification of images. From the study of attention-driven image pre-classification and retrieval, we know that images can be classified into two types. For the images with distinct objects, we call them attentive ones. On the contrary, for the images which do not contain distinct objects, we call them non-attentive images. Furthermore, attention-driven strategy is able to extract important objects from the attentive images and retrieve this type of images efficiently. Besides, fusing-all strategy, which extracts all objects from an image and gives equal weights to all objects, is utilized to retrieve non-attentive images. Previous researchers always paid more attention to retrieval strategies instead of the pre-classification stage of the image database. In their system, the query images will be compared with the whole database, without considering the classification of the database images. If the database contains a great number of images, the retrieval strategies may take a lot of time and lead to poor retrieval results. Therefore, an neural network based pre-classification procedure is adopted to the image database before retrieval, so as to increase the retrieval efficiency and reduce the retrieval error Furthermore, besides attentive and non-attentive classes, there are some special images which can not be classified into a certain class even by manual classification. So we define a new image class named "unsure", which means we can not justify these images easily. At last, we evaluate the efficiency and performance improvements in image retrieval stage after the proposed database pre-classification.

Files in this item

Files Size Format
b23059400.pdf 3.298Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

     

Quick Search

Browse

More Information