ECCV 2012 - LNCS 7572-7578 and 7583-7585

Online Video Segmentation by Bayesian Split-Merge Clustering

Juho Lee¹, Suha Kwak¹, Bohyung Han^{1, 3}, and Seungjin Choi^{1, 2, 3}

¹Department of Computer Science and Engineering, Pohang University of Science and Technology, 77 Cheongam-ro, Nam-gu, Pohang 790-784, Korea
stonecold@postech.ac.kr
mercury3@postech.ac.kr
bhhan@postech.ac.kr
seungjin@postech.ac.kr

²Division of IT Convergence Engineering, Pohang University of Science and Technology, 77 Cheongam-ro, Nam-gu, Pohang 790-784, Korea

³Department of Creative IT Excellence Engineering, Pohang University of Science and Technology, 77 Cheongam-ro, Nam-gu, Pohang 790-784, Korea

Abstract. We present an online video segmentation algorithm based on a novel nonparametric Bayesian clustering method called Bayesian Split-Merge Clustering (BSMC). BSMC can efficiently cluster dynamically changing data through split and merge processes at each time step, where the decision for splitting and merging is made by approximate posterior distributions over partitions with Dirichlet Process (DP) priors. Moreover, BSMC sidesteps the difficult problem of finding the proper number of clusters by virtue of the flexibility of nonparametric Bayesian models. We naturally apply BSMC to online video segmentation, which is composed of three steps—pixel clustering, histogram-based merging and temporal matching. We demonstrate the performance of our algorithm on complex real video sequences compared to other existing methods.

LNCS 7575, p. 856 ff.

Full article in PDF | BibTeX