Output details
11 - Computer Science and Informatics
University of Bath
Finding semantic structures in image hierarchies using Laplacian graph energy
<23>Hierarchical image segmentations are often overly complex. Our method reduces complexity by an order of magnitude without information loss. Published in ECCV, a highly ranked Computer Vision conference with an acceptance rate around 20%, this work provided a foundation for our later work with tracking (Neurocomputing 2012), object recognition (BMVC 2012,BMVC 2013), and non-photorealistic rendering (TVCG 2013,CAe 2013). The method is now in every day use in our EPSRC project ”Classifying Images Regardless of Depictive Style”. It was developed jointly with UC Berkeley, and is routinely used there too. It has also influenced research at Kyushu University (Perera’s PhD thesis).