Paper
27 January 2009 Motion classification for image stabilization with moving objects
Author Affiliations +
Abstract
Feature based image stabilization is often undermined by moving object. A new method to classify motion vector based on K-Medoids cluster is proposed in this paper for removing the influence of moving objects in the scene while image stabilizing. In the proposed method, the image is divided into several small areas before features being extracted from the regions firstly. Features are tracked by finding the corresponding ones in the next frame along the image sequence. Then a K-Medoids cluster method is proposed to classify those features by their motion vector to distinguish them from background or from moving objects. A dispersivity value of internal distance among those features in one class is defined to confirm that which cluster is from background. The experiment shows that the result is good to remove the features from moving objects. The limitation number of total feature and cluster assures the running speed fast enough for real-time processing on PC.
© (2009) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Yubin Zhou "Motion classification for image stabilization with moving objects", Proc. SPIE 7156, 2008 International Conference on Optical Instruments and Technology: Optical Systems and Optoelectronic Instruments, 71561F (27 January 2009); https://doi.org/10.1117/12.805544
Advertisement
Advertisement
RIGHTS & PERMISSIONS
Get copyright permission  Get copyright permission on Copyright Marketplace
KEYWORDS
Feature extraction

Image classification

Cameras

Image processing

Motion models

Motion estimation

Digital filtering

Back to Top