PileBars: scalable dynamic thumbnail bars

Brivio Paolo, Ponchio Federico, Tarini Marco, Cignoni Paolo

Thumbnails; Browsing; Images 

We introduce PileBars, a new class of animated thumbnail-bars aimed at easing the browsing of very large image datasets (thousands of images). Since the bar is meant to be just one element of a GUI, it covers only a small portion of the screen; yet it provides a global view of the entire dataset, without any scrolling panels. Instead, thumbnails are dynamically rearranged, resized and reclustered into adaptive layouts during the entire browsing process. The objective is to make the user able, at the same time, to accurately pinpoint a specific image (even among semantically close ones), and to jump anywhere to "distant" parts of the dataset. The used thumbnail layouts also maximize temporal coherence, thus allowing for smooth animations from one layout to the next. The system is very general: it can be driven by any application-specific image-to-image semantic distance function, and respects any user-defined total ordering of the images; the ordering can be either inferred from the semantic or be chosen independently from it, depending on the application. The applicability of the resulting system is tested in a number of practical applications: browsing image collections from the web or personal achieves; navigation of 3D virtual scenes based on calibrated pictures; selection of a frame in a movie. In each test scenario, proper distance and ordering functions are fed to the system.

Source: ISTI Technical reports, 2011