The Web contains wide variety of images published by the general public. Thus, it is difficult for conventional image retrieval methods to retrieve images from the Web by specifying image contents precisely. Most of current search users first retrieve Web pages (or parts of Web pages) that contain related topics to the target image, then browse the Web pages for discovering the images with appropriate contents. This gap causes problems on search-ability and reliability of image retrieval. In order to fill the gap, we propose an approach for retrieving “Web contexts”of images, each of which represents with what kind of Web contents (e.g. texts and other images) an image is associated. A Web context is represented by a set of Web contents surrounding the image. We define three types of Web contexts for an image in accordance with different associations between an image and Web contents based on Web components (i.e. structured documents and hyperlinks). Our approach enables a search user to retrieve images by their usages directly from the Web. Besides, by showing each of result images as a part of the corresponding Web context, a search user can visually understand relevancy between query request and the result image. In this paper, we propose methods for extracting and visualizing Web contexts of images.
[1]
Michael McGill,et al.
Introduction to Modern Information Retrieval
,
1983
.
[2]
Mark D. Dunlop,et al.
Image retrieval by hypertext links
,
1997,
SIGIR '97.
[3]
Dragutin Petkovic,et al.
Query by Image and Video Content: The QBIC System
,
1995,
Computer.
[4]
Ehud Rivlin,et al.
Placing search in context: the concept revisited
,
2002,
TOIS.
[5]
H. P. Luhn.
Key word‐in‐context index for technical literature (kwic index)
,
1960
.
[6]
Shih-Fu Chang,et al.
Querying by color regions using VisualSEEk content-based visual query system
,
1997
.
[7]
Steve Lawrence,et al.
Context in Web Search
,
2000,
IEEE Data Eng. Bull..