Video Segmentation via Adaptive Higher-Order CRF with Windowed Dynamics

In this paper, we propose an adaptive higher-order CRF (HCRF) labeling approach towards automatic video object segmentation with better performances, higher effectiveness and efficiency. In comparison with the existing state of the arts, our approach achieves further improvements in terms of segmentation results. Our contribution can be highlighted as: (i) HCRF labeling is made adaptive to video content changes via a windowed dynamics; (ii) Fusion of multiple features is automatically optimized via adaptive weight parameters; (iii) Our approach is suitable for pre-segmentations base on a simpler image segmentation to reduce the overall computing cost, and be easily generalized without compromising on its performances. The experimental results show that our method can improve the video segmentation performance.

[1]  Pushmeet Kohli,et al.  Robust Higher Order Potentials for Enforcing Label Consistency , 2008, 2008 IEEE Conference on Computer Vision and Pattern Recognition.

[2]  Mei Han,et al.  Efficient hierarchical graph-based video segmentation , 2010, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[3]  Eric L. Miller,et al.  Multiple Hypothesis Video Segmentation from Superpixel Flows , 2010, ECCV.

[4]  Hsu-Yung Cheng,et al.  Region segmentation and labeling in aerial surveillance applications , 2012, 2012 12th International Conference on ITS Telecommunications.

[5]  Pascal Fua,et al.  SLIC Superpixels Compared to State-of-the-Art Superpixel Methods , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[6]  Andrew McCallum,et al.  Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data , 2001, ICML.

[7]  Narendra Ahuja,et al.  Exploiting nonlocal spatiotemporal structure for video segmentation , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[8]  Vladlen Koltun,et al.  Parameter Learning and Convergent Inference for Dense Random Fields , 2013, ICML.

[9]  Luis Jiménez,et al.  Real-time moving object segmentation in H.264 compressed domain based on approximate reasoning , 2009, Int. J. Approx. Reason..

[10]  Ling Shao,et al.  Human action segmentation and recognition via motion and shape analysis , 2012, Pattern Recognit. Lett..

[11]  Hsu-Yung Cheng,et al.  Segmentation of Pedestrians with Confidence Level Computation , 2013, J. Signal Process. Syst..