An innovative content-based indexing technique with linear response suitable for pervasive environments

In the world of pervasive computing the status of each device (idle or active) could be dynamically altered and multimedia content is added or removed dynamically. The traditional approaches that create static multimedia indices are inappropriate. To address this problem, a new technique for indexing multimedia content in pervasive environments, is proposed. This technique is based on an innovative algorithm that indexes the multimedia content to a uniform representation (M-hyper rectangle), using a cell topology structure. Groups (clusters) with images that share common content are created using the aforementioned cell topology. The resulting representation is used in order to yield a significant reduction of the computational complexity and retrieval response time. More specifically, upon an image query, the groups whose content is the closest to the query are selected and on a later step, the closest images to the query are retrieved from the selected groups and presented to the end user. All comparisons among the stored and query data, as well as the construction of the underlying cluster are performed using Boolean operations, making this method suitable for a highly dynamic environment.