Multimedia Data Indexing

Multimedia Data Indexing And Management

Media data like images, video clips, audio clips are often represented by their d-dimensional features, and each media object becomes a point in some feature space R^d. How to find a proper feature representation that captures necessary information for effective search and retrieval tasks is the active research topic in the multimedia analysis community. Industrial standard like MPEG-7 also captures the state of the art of image/video features and metrics for search and retrieval, including color, shape and texture features for images, as well as motion and camera features for video.

The purpose of multimedia data base system is to have structured storage and indexing such that the amount of I/O operations that load data into memory, and also the amount computation involved for a query request can be reduced and therefore achieving fast response time to end users. This is also a well studied field dates back to early days of computer science and with a rich set of tools and solutions. The multimedia data base present unique challenges in the sense that the typical feature data space dimensionality is high, and if this issue is not dealt with effectively, the retrieval performance of the traditional indexing and storage solutions degenerates quickly.

In this section, we discuss the multimedia data indexing and storage solutions, and the typical queries supported. The effect of dimensionality on retrieval performance is discussed in detail.