A semi-supervised video object segmentation algorithm using multiple random walkers (MRW) is proposed in this work. We develop an initial probability estimation scheme that minimizes an objective function to roughly separate the foreground from the background. Then, we simulate MRW by employing the foreground and background agents. During the MRW process, we update restart distributions using a hybrid of inference restart rule and interactive restart rule. By performing these processes from the second to the last frames, we obtain a segment track of the target object. Furthermore, we optionally refine the segment track by performing Markov random field optimization. Experimental results demonstrate that the proposed algorithm significantly outperforms the state-of-the-art conventional algorithms on the SegTrack v2 dataset.
|Publication status||Published - 2016|
|Event||27th British Machine Vision Conference, BMVC 2016 - York, United Kingdom|
Duration: 2016 Sept 19 → 2016 Sept 22
|Other||27th British Machine Vision Conference, BMVC 2016|
|Period||16/9/19 → 16/9/22|
Bibliographical noteFunding Information:
This work was supported partly by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. NRF-2015R1A2A1A10055037), and partly by the MSIP, Korea, under the ITRC support program supervised by the Institute for Information &communications Technology Promotion (No. IITP-2016-R2720-16-0007).
This work was supported partly by the National Research F oundation of K orea (NRF) grant funded by the K orea go v ernment (MSIP) (No. NRF-2015R1A2A1A10055037), and partly by the MSIP , K orea, under the ITRC support program supervised by the Institute for Infor - mation &communications T echnology Promotion (No. IITP-2016-R2720-16-0007).
© 2016. The copyright of this document resides with its authors.
ASJC Scopus subject areas
- Computer Vision and Pattern Recognition