Please use this identifier to cite or link to this item: http://scholarbank.nus.edu.sg/handle/10635/13773
Title: Featureless techniques for estimating camera motion (scale and rotation) in videos
Authors: ABHINAV SINGH
Keywords: Camera Motion Reconstruction, Motion Detection and Estimation, Content Based Indexing of Video, Computer Vision, Feature Tracking, Video Manipulation
Issue Date: 5-May-2004
Source: ABHINAV SINGH (2004-05-05). Featureless techniques for estimating camera motion (scale and rotation) in videos. ScholarBank@NUS Repository.
Abstract: This thesis presents two featureless techniques for efficient and robust reconstruction of camera motion from digital video. Existing approaches, if reliable, are slow and computation intensive; or too unreliable and approximate, if fast. The proposed techniques present a good blend of computational efficiency and quality, with results acceptable to most practical applications.The first technique relies upon the Phase-Correlation method to derive motion field from videos, while the second one takes MPEG motion-vectors as approximate representations of this field. These motion fields are then thoroughly processed and matched with geometrical models to derive camera parameters.The strength of our approach lies in the emphasis on proposed refinement steps and robust object-background segmentation a?? which ensure that the reconstruction is of comparable quality, despite the relatively coarse nature of the motion fields, to a more sophisticated technique that uses dense sub-pixel optical flow motion estimation method. Substantial speed gains are also demonstrated.
URI: http://scholarbank.nus.edu.sg/handle/10635/13773
Appears in Collections:Master's Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
[MSc_Thesis]_Abhinav_Singh_2003.pdf4.2 MBAdobe PDF

OPEN

NoneView/Download

Page view(s)

281
checked on Dec 11, 2017

Download(s)

169
checked on Dec 11, 2017

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.