Logo image
Back
Automatic video object segmentation from VOP
Conference paper   Open access

Automatic video object segmentation from VOP

F.A. Sohel, C.M. Rahman and G.C. Karmakar
International Conference on Electrical, Electronics and Computer Engineering (ICEECE-03) (Dhaka, Bangladesh, 22/12/2003–24/12/2003)
2003
pdf
vop_paper_ieece_version2.pdfDownloadView
Open Access

Abstract

The video coding standard MPEG-4 is enabling content-based functionalities of a prior decomposition of sequences into video object planes (VOP) so that each VOP represents a semantic object. Therefore extraction of semantic objects is an important part. There are various coding tools: shape coding, motion estimation and compensation, texture coding, multifunctional coding, error resilience, sprite coding and scalability. These are performed by using diverse techniques, like: pixel based segmentation, region based segmentation, boundary based segmentation, morphological segmentation, Bayesian segmentation, model based approaches etc. However, most of the techniques are either manual or semi-automatic. Semiautomatic methods require user assistance and suffer from addressing the following issues: the background is variable due to the camera motion, the light condition slightly changed, new object can appear at any time, objects may remain for a long time in the scene, many objects may be in a scene and possible occlusion. So it is very important to develop automatic techniques that are robust and fast and will combine low-level automatic feature segmentation with interactive method for defining and tracking high semantic video objects and also the above mentioned constraints. In this paper we have proposed how this might be done. It can be done in accordance with the following modules: sprite generation from background and statistical feature analysis approach for object definition, region based motion estimation for tracking.

Details

Metrics

84 File views/ downloads
56 Record Views
Logo image