Share Email Print
cover

Proceedings Paper

Adaptive segmentation technique for automatic object region and boundary extraction for activity recognition
Author(s): Fatema A. Albalooshi; Vijayan K. Asari
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

An unsupervised adaptive segmentation technique based on active contour model for automatic object region and boundary extraction in video sequences is proposed. The active contour model identies each region using certain region descriptors that guide the motion of the initial contour towards the actual region of interest. The region descriptors depend on the statistical information inside and outside the initial contour, which control the growth or contraction of the contour. A process of selective binary and Gaussian ltering regularization level set smooths the level-set function in order to maintain the level of expansion of the active contour. The adaptivity of the proposed method comes from automatic selection of an optimal Gaussian lter parameter which depends on the characteristics of the input video frames. More precisely, the algorithm utilizes the statistical information of each video frame to identify the optimal value of the Gaussian parameter which should be large enough to cover most of the level-set function, but not too large to overlap multiple neighbouring objects. The optimal value of Gaussian parameter aects the segmentation performance of the entire activity video sequence. The performance of the proposed method was tested and evaluated on Weizmann database and KTH database which represent object regions in dierent illuminations and scenarios. The performance of the proposed method is also compared to several state-of-the-art image segmentation methods and observed improved performance in terms of accuracy and eciency. The study demonstrates the robustness of the method to non-uniform illuminations and background noises.

Paper Details

Date Published: 7 May 2012
PDF: 11 pages
Proc. SPIE 8399, Visual Information Processing XXI, 839907 (7 May 2012); doi: 10.1117/12.919187
Show Author Affiliations
Fatema A. Albalooshi, Univ. of Dayton (United States)
Vijayan K. Asari, Univ. of Dayton (United States)


Published in SPIE Proceedings Vol. 8399:
Visual Information Processing XXI
Mark Allen Neifeld; Amit Ashok, Editor(s)

© SPIE. Terms of Use
Back to Top