Research And Implementation On Photo And Video Collage Composition

The dramatic growth of digital photo and video data requiresrepresentation technique which supports not only fast but also pictorialbrowsing of content.With the explosion of image files number, the ability tomanage the content of a huge number of images has become a key skill.One of the most effective and visually appealing techniques of imagebrowsing is representation based on collage. A collage represents an imagelibrary or video material as a set of meaningful snapshots that allowssummarizing the content of the library.Automatic collage design is a widely spread task in many applications,from a design of home collages to commercial advertising projects.In our research we proposed two approaches of automatic collagecomposition: ROI Shape Estimation Based Approach and Adaptive ROI BlockAnalysis Based Approach.Originally proposed ROI Shape Estimation Based Approach has resultsof composition performance, informative and aesthetic values very close tointernational existing techniques.Adaptive ROI Block Analysis Based Approach is better than existingtechniques. According to series of experimental analysis and social networkbased surveys (estimated by up to300respondents) its average time ofcomplete cycle of collage composition with frame selection, allocation andROI approximation–17.98sec. Diversity of generated layouts is81.16%with stable0%level of empty areas; accuracy of ROI approximation is86.08%; information value and aesthetic value of entire output collage areestimated as90.25%and90.83%respectively.Adaptive ROI Block Analysis Based Approach about4seconds fasterthan Autocollage and produces output collages with5%better informationvalue and10%better aesthetic value.Our approach is proposed for managing both photo libraries and videomaterial and compilation more spectacular output collages.