Abstract—This paper presents the implementation of web
image retrieval system using keyword-based search and visual
image features. In order to correctly correlate terms to a web
image, the associated text of the web image is partitioned into
text blocks according to the structure of the text with respect to
the web images. Then, keywords are extracted and stored in
image indexing database which will later be used in keyword
based retrieval. When user enters keyword, result images are
generated by image indexing and searching algorithms. Most of
the web image search systems are based on only keyword based
searches. But most of the result images in the keyword search
are not relevant to the query. To reduce the influence of those
irrelevant images, visual image features are used. Firstly, image
features are extracted and then they are stored in the image
indexing database. And then these features are used to cluster
images for relevant and non-relevant. Combination of keyword
search and visual feature extraction will result in producing
more relevant images by removing non-relevant images.
Extracting visual features can improve the system performance.
Index Terms—Image retrieval system, keyword-based search,
visual image features.
Nyein Myint Myint Aung is with University of Technology, Yatanarpon
Cyber City, Myanmar (e-mail: thae.thae.star@gmail.com).
Cite: Nyein Myint Myint Aung, "Combination of Keyword and Visual Features for Web Image Retrieval System," International Journal of Information and Education Technology vol. 4, no. 6, pp. 487-490, 2014.