Hierarchical Navigation and Visual Search for Video Keyframe Retrieval Carles Ventura, Manel Martos, Xavier Gir´ o-i-Nieto, Ver´ onica Vilaplana, and Ferran Marqu´ es Technical University of Catalonia (UPC), Barcelona, Catalonia / Spain, {carles.ventura,xavier.giro,veronica.vilaplana,ferran.marques}@upc.edu Abstract. This work presents a browser that supports two strategies for video browsing: the navigation through visual hierarchies and the retrieval of similar images and objects. The input videos are firstly pro- cessed by a keyframe extractor to reduce the temporal redundancy and decrease the number of elements to consider. These generated keyframes are hierarchically clustered with the Hierachical Cellular Tree (HCT) algorithm, an indexing technique that also allows the creation of data structures suitable for browsing. Different clustering criteria are avail- able, in the current implementation, based on four MPEG-7 visual de- scriptors computed at the global scale. The navigation can directly drive the user to find the video timestamps that best match the query or to a keyframe which is globally or locally similar in visual terms to the query. If this is the case, a visual search engine is also available to find other similar keyframes or regions, also based on MPEG-7 visual descriptors. Keywords: Video browser, Hierarchical Navigation, Object retrieval, Image retrieval 1 Introduction As a consequence of recent technology development, large amounts of video data are generated and stored. Accessing these rich portion of data in terms of au- diovisual and semantic content is an open research issue with several solution depending on the user needs. This work proposes an hybrid interface that com- bines both navigation and search tools to provide users with a high degree of flexibility for choosing the most appropriate strategies according to the query nature. Figure 1 shows a screenshot of GOS, the GUI that exploits the presented techniques. 2 Hierarchical Cellular Tree The Hierarchical Cellular Tree (HCT) algorithm [1] was designed to bring an effective solution for indexing large multimedia databases. The elements are partitioned depending on their relative distances and stored within cells on the