Accuracy enhancement of function-oriented web image classification

Authors:
Koji Nakahira;Toshihiko Yamasaki;Kiyoharu Aizawa
Affiliations:
The University of Tokyo, Kashiwano-ha, Kashiwa-shi, Chiba, Japan;The University of Tokyo, Kashiwano-ha, Kashiwa-shi, Chiba, Japan;The University of Tokyo, Kashiwano-ha, Kashiwa-shi, Chiba, Japan
Venue:
WWW '05 Special interest tracks and posters of the 14th international conference on World Wide Web
Year:
2005

Citing 4
Cited 0

Discrete cosine transform: algorithms, advantages, applications

Discrete cosine transform: algorithms, advantages, applications
Support-Vector Networks

Machine Learning
Visually Searching the Web for Content

IEEE MultiMedia
Automatic categorization for WWW images with applications for retrieval navigation

PCM'04 Proceedings of the 5th Pacific Rim conference on Advances in Multimedia Information Processing - Volume Part I

Quantified Score

Hi-index	0.00

Visualization

Abstract

We propose a function-oriented classification of web images and show new applications using this categorization. We defined nine categories of images taking into account of their functions used in web pages, and classified web images by using Support Vector Machine (SVM) in tree structured way. In order to achieve high accuracy of classification, we employed two kinds of features, image-based features and text-based features, and the two kinds can be used together or separately for the stages of the classification. We also utilized DCT coefficients to classify photo images and illustrations. As a result, accurate classification has been achieved. Finally, we show the page summarization as a new application that is made feasible for the first time by our new categories of WWW images.