Title: Video stabilisation using local salient feature in particle filter framework

Authors: Guangju Chen; Zhiqiang Ma; Yong Shan; Xiaoyan Zhang

Addresses: Department of Network Engineering, Institute of Telecommunication Engineering, Air Force Engineering University, Xi'an, Shaanxi, 710077, China ' Department of Network Engineering, Institute of Telecommunication Engineering, Air Force Engineering University, Xi'an, Shaanxi, 710077, China ' Department of Network Engineering, Institute of Telecommunication Engineering, Air Force Engineering University, Xi'an, Shaanxi, 710077, China ' Department of Network Engineering, Institute of Telecommunication Engineering, Air Force Engineering University, Xi'an, Shaanxi, 710077, China

Abstract: A novel digital video stabilisation approach that uses salient features and particle filter for global motion estimation is proposed in this paper. In this approach, the local salient features are first gained by the improved K-means clustering method using the features obtained from the Scale Invariant Feature Transform (SIFT) feature points extracted from the down-sampled images, and then the salient regions are located in the video frames. The trajectory of the features extracted from the salient regions is used to estimate global motion between frames. A new similarity function called SRMSE (Salient Region Mean Square Error) is proposed in particle filter framework to reduce the computational cost. Motion compensation yields stabilised video sequences using the estimates. Experimental results demonstrate that the proposed algorithm has good performance and improves the efficiency significantly.

Keywords: SIFT; scale invariant feature transform; k-means clustering; salient features; salient region; particle filter; video stabilisation; global motion estimation; motion compensation.

DOI: 10.1504/IJWMC.2013.053034

International Journal of Wireless and Mobile Computing, 2013 Vol.6 No.1, pp.72 - 79

Received: 31 May 2012
Accepted: 03 Jul 2012
Published online: 11 Oct 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article