Efficient Optimization of Photo Collage

Yichen Wei,   Yasuyuki Matsushita, Yingzhen Yang
Visual Computing Group State Key Lab of CAD & CG
Microsoft Research Asia Zhejiang University

Abstract

Photo collage is an effective representation for photo summarization and visualization. However, its practical usage is limited by its high computational cost. In this paper, we present efficient optimization techniques based on a novel formulation on markov random fields and several insights unexploited by previous approaches. Our method achieves up to hundreds-fold performance improvement, and user/application specific constraints can be easily integrated. We present new applications including interactive collage refinement and dynamic collage for photo browsing. Preliminary evaluation and user study indicates that our approach extends the usability of photo collage.

Publication

Technical Report, Microsoft Research£¬MSR-TR-2009-59, May, 2009, pdf 2M

A preliminary version of this work is published on The Visual Computer, vol.25, pp. 431-439, March, 2009, pdf 2M

Demo

Interactive Collage(wmv, 4M), Dynamic Collage(wmv, 20M)

Results

initial result after swapping two faces
Interactive collage allows a user to edit and update result in a responsive time. In this example, a user finds the initial result (left) unsatisfactory since the same faces are grouped together (red circle). The user then swaps the two faces (green arrow) and obtains the updated result (right) within a second.


Dynamic collage achieves temporally coherent and spatially compact photo browsing/visualiazation experience. It is dynamically updated by replacing old photos with new ones. It is suitable to browse a large number of photos, such as web images or family album.