Intelligent scissors for image composition
SIGGRAPH '95 Proceedings of the 22nd annual conference on Computer graphics and interactive techniques
Mean Shift: A Robust Approach Toward Feature Space Analysis
IEEE Transactions on Pattern Analysis and Machine Intelligence
What Energy Functions Can Be Minimizedvia Graph Cuts?
IEEE Transactions on Pattern Analysis and Machine Intelligence
ACM SIGGRAPH 2004 Papers
"GrabCut": interactive foreground extraction using iterated graph cuts
ACM SIGGRAPH 2004 Papers
An Experimental Comparison of Min-Cut/Max-Flow Algorithms for Energy Minimization in Vision
IEEE Transactions on Pattern Analysis and Machine Intelligence
Interactive image segmentation by maximal similarity based region merging
Pattern Recognition
Enhancing interactive image segmentation with automatic label set augmentation
ECCV'10 Proceedings of the 11th European conference on Computer vision: Part VI
Interactive segmentation with super-labels
EMMCVPR'11 Proceedings of the 8th international conference on Energy minimization methods in computer vision and pattern recognition
Active learning for semantic segmentation with expected change
CVPR '12 Proceedings of the 2012 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
Hi-index | 0.00 |
Compared to automatic segmentation, interactive segmentation is a flexible method to separate the interesting object from background. However, satisfactory results may not be achieved even with lots of interactions since user's operation may not provide enough information to decide the labels of ambiguous regions. To deal with this problem, we present an interactive segmentation approach based on active learning scheme, which can automatically recommend the most informative regions to guide the user interactions. Our method employs a two-step strategy. Firstly, based on initial user interactions, it adopts active learning to iteratively select the most crucial regions and query the oracle for their true labels. In the second step, we minimize an energy function, which combines low-level features extracted from total interactions, to segment the object. Experimental results demonstrate our method can achieve high segmentation accuracy within desirable interactions.