Category Independent Object-Level Saliency Detection
Title | Category Independent Object-Level Saliency Detection |
Publication Type | Conference Paper |
Year of Publication | 2013 |
Authors | Jia, Y., & Han M. |
Other Numbers | 3618 |
Abstract | It is known that purely low-level saliency cues such asfrequency does not lead to a good salient object detectionresult, requiring high-level knowledge to be adopted forsuccessful discovery of task-independent salient objects. Inthis paper, we propose an efficient way to combine suchhigh-level saliency priors and low-level appearance models. We obtain the high-level saliency prior with the objectness algorithm to find potential object candidates withoutthe need of category information, and then enforce the consistency among the salient regions using a Gaussian MRFwith the weights scaled by diverse density that emphasizesthe influence of potential foreground pixels. Our model obtains saliency maps that assign high scores for the wholesalient object, and achieves state-of-the-art performance onbenchmark datasets covering various foreground statistics. |
URL | https://www.icsi.berkeley.edu/pubs/vision/categoryindependent13.pdf |
Bibliographic Notes | Proceedings of the International Conference on Computer Vision 2013 (ICCV 2013), Sydney, Australia |
Abbreviated Authors | Y. Jia and M. Han |
ICSI Research Group | Vision |
ICSI Publication Type | Article in conference proceedings |