In the information era, the old adages of ''seeing is believing'' and ''a picture is worth a thousand words'' can be attested to and realized more than ever, thanks to the advent of enabling technologies such as broadband networks and internet, high-powered workstations, multimedia databases, digital image and video compression standards, and VLSI technology. There are a growing number of applications including video-on-demand, video over internet, interactive television, digital libraries, video games, etc., which extensively use the visual media. A key requirement in these applications is efficient indexing, storage, retrieval, and browsing of images and video.
Researchers around the world are forging ahead at a rapid pace in this crucial area of visual computing and communications. The papers appearing in this special issue not only represent some of the latest results on indexing, storage, browsing, and retrieval in both the uncompressed and compressed domains, but also provide a comprehensive overview of this exciting field. It is hoped that this issue will facilitate further exploration of this subject.
The first two papers deal with the key aspect of indexing using spatial relationships between objects. Zhang, Chang, and Yau propose a unified iconic indexing scheme for images in visual databases. The paper by Shearer, Venkatesh, and Kieronska presents a system for indexing video sequences based on the spatial relationships of the objects extracted from the sequence.
The specification and extraction of visual content is crucial for efficient indexing, browsing and retrieval. The next three papers propose techniques for extracting the content from video. The paper by Kim presents an automatic text detection and location method and its application to content-based retrieval. The paper by Pfeiffer, Leinhart, Fischer, and Effelsberg provides a methodology for automatically abstracting digital videos based on the content characteristics. Sudhir and Lee propose an approach to annotating video sequences by extracting the dominant camera motion using optical flow streams.
Retrieval by content from multimedia databases is becoming increasingly popular