MSc Semester Project
Image tag propagation
Prof. Dr. Touradj Ebrahimi
Assistant: Ivan Ivanov
June 11, 2010
Over the last few years, social networks have greatly increased Internet users involvement in content creation and annotation. These systems are characterized as easy-to-use and interactive. Users contribute with their opinion by annotating content (the so-called tagging), they add tags, comments, and recommendations or rate these content.
This semester project aims at developing a system to enrich images with social network tagging. In particular, it plans to exploit different visual features so as to extract the content from an image and to use them for efficient image annotation. More specifically, the user provides an image, which he/she tags. The system performs image similarity search in order to find images having the same content from a large collection of images. Initial tags of the initial image can then be propagated to other images in the large collection. For instance, a tourist takes a picture of the Eiffel tower with his mobile phone. By querying Flickr, the system can identify a number of images also depicting the Eiffel tower. The provided tags, e.g. “Eiffel tower”, “Paris”, are merged and finally assigned to other images in the dataset. This also allows determining the context of the picture. This approach offers a compelling new look at how metadata can be easily generated in order to provide efficient context-aware management and organization of image collections.
More specifically, the goal of this project is to study different approaches for content-based image retrieval, to assess and to compare their performance for tag propagation in still images. The following tasks should be performed:
- Study different approaches for content-based image retrieval, different local and global features used for image description,
- Experiment with dominant color, edge and texture, as well as SIFT features,
- Implement and validate a tag propagation system developed in software, which takes a query image as input, retrieves similar images from a database ranked by their degree of similarity, and propagates tags of the query image to those images that are relevant relying on a user feedback mechanism,
- Assess and compare the performance of different features used in the above system.