Related Experiment Videos
A Query Expansion Framework in Image Retrieval Domain Based on Local and Global Analysis
M M Rahman1, S K Antani, G R Thoma
1U.S. National Library of Medicine, National Institutes of Health, Bethesda, MB, USA.
Information Processing & Management
|August 9, 2011
Summary
This study introduces an image retrieval system using concept features and automatic query expansion. The framework enhances search precision and recall by analyzing concept correlations.
Area of Science:
- Computer Science
- Information Retrieval
- Image Analysis
Background:
- Traditional text retrieval uses keyword vectors.
- Image retrieval often lacks robust concept representation and query expansion.
Purpose of the Study:
- To develop an image retrieval framework using automatic query expansion in a concept feature space.
- To generalize the vector space model for image retrieval.
Main Methods:
- Images represented as vectors of weighted concepts ('bag of concepts') derived from color and texture patches.
- Support Vector Machine (SVM) classification used for concept vocabulary generation.
- Query expansion techniques based on local (co-occurrence, neighborhood proximity) and global (similarity thesaurus) concept analysis.
Main Results:
- Demonstrated effectiveness on natural scenes and biomedical image databases.
- Improved precision and recall in image retrieval tasks.
- Addressed feature independence assumption via concept correlation analysis.
Conclusions:
- The proposed framework offers an effective approach to image retrieval.
- Automatic query expansion in concept feature space significantly enhances retrieval performance.
- The method is applicable across diverse image collections.