We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
A framework for automated image annotation.
- Authors
Wang, L.; Khan, L.; Thuraisingham, B
- Abstract
In this paper we propose a weighted feature selection algorithm and apply it to automatic image annotation and retrieval. For automate image annotation, we generate a finite set of visual tokens using clustering, and learn the correlation between keywords and visual tokens. Next, this learned model is used for the annotation of images that do not have captions. During the clustering process for a given cluster we determine relevant features based on conservative and aggressive histogram analysis, and assign greater weight to relevant features compared to less relevant features. We have implemented various models to link visual tokens with keywords based on the clustering results of a K-means algorithm, with weighted feature selection and without feature selection, and evaluated performance using precision, recall and correspondence accuracy using a benchmark dataset. The results show that weighted feature selection is better than traditional selection methods for automatic image annotation and retrieval, and that conservative weighted feature selection is better than aggressive feature selection.
- Subjects
ALGORITHMS; IMAGING systems; CLUSTER analysis software; INTERNET programming; BENCHMARKING (Management); INFORMATION retrieval; COMPUTER network resources
- Publication
Computer Systems Science & Engineering, 2007, Vol 22, Issue 1/2, p15
- ISSN
0267-6192
- Publication type
Article