Tobun D. Ng, Michael G. Christel, Alexander G. Hauptmann, and Howard D. Wactlar
The video collage is a novel and effective interface for dynamically summarizing and presenting mined multimedia information from video collections. It provides a flexible mechanism for inferring knowledge and trailing analytic thoughts, which are the essential tasks in multimedia knowledge management. We will describe multimedia knowledge management, discuss how collages are automatically produced, illustrate their use for multimedia knowledge management, and evaluate their effectiveness as summaries across news stories. Collages are presentations of text and images extracted from multiple video sources. They provide an interactive visualization for a set of analyzed video documents, summarizing their contents and offering a navigation aid for further exploration. The dynamic creation of collages is based on user context, e.g., an originating query, coupled with automatic processing to refine the candidate imagery. Named entity identification and common phrase extraction provides associative textual description. The dynamic manipulation of collages allows user-directed filtering as well as revealing additional details and previously undiscovered content. The utility of collages as summaries is examined with respect to other published news summaries.