Epitome (data processing): Difference between revisions

Revision as of 07:40, 25 June 2008

Revision as of 19:56, 20 August 2013

Line 1:

In [[image processing]], an '''epitome''' is a condensed digital representation of the essential [[statistical]] properties of ordered [[dataset]]s, such as matrices representing [[image]]s, [[audio signal]]s, [[video]]s, or [[genetic sequence]]s. Although much smaller than the epitomized data, the epitome contains many of the smaller overlapping parts of the data with much less repetition and with some level of generalization. As such, it can be used for data mining and other machine learning and signal processing tasks. The first use of '''epitomic analysis''' was performed on image textures<ref>Nebojsa Jojic's image epitome website [http://www.research.microsoft.com/~jojic/epitome.html].</ref> and was used for image parsing. The epitome model has also been applied to videos.<ref>Vincent Cheung's video epitome website [http://www.psivincentcheung.toronto.educa/~vincentresearch/videoepitome.html].</ref> Filling in missing parts of a video, removing objects from a scene and performing video [[super-resolution]] are examples of tasks in which the video epitome has proven useful.

Epitomes are also being investigated as tools for [[rational vaccine design]].