Context-based image retrieval in Fronter learning environment

Abstract

The Internet has become a natural medium for finding information and resources, and has probably become the most important tool in education and e-learning as well. Many educational institutions use on-line systems for uploading, creating and publishing educational content to students and pupils. Extended use of multimedia files, video, audio and image, as a part of the content is a growing trend and there is ever more a need to search for desired multimedia content. This causing challenges to both the on-line systems and its users. To make multimedia content suited for search and retrieval it is imperative to organize and describe content well. Normally, users do not spend much of their time annotating and organizing content. Text-based search engines that are integrated into the educational on-line systems are normally not very suitable for search of multimedia content. The specific concern of this project is to investigate and suggest solutions to how image context information can be collected and then used in image annotation and retrieval within educational content. The image context in this setting is defined as the image environment in which the image is used. For example it can be a collection of images or documents, the course(s), subtopic(s) and/or assignments where images are used as illustrations. The system that is designed and implemented in this thesis will use image context to describe the images content and its semantics and use this in its image retrieval. If there are multiple versions of an image it will also be used in the description and retrieval processes. The evaluation of the system indicates that usage of the image context is very useful in describing the image content and its semantics. The system finds and retrieves more relevant images to the search than the integrated search engine in on-line educational portal. It has also good retrieval performance compared to the integrated one.