Analysis of planar‐motion segmentation using affine fundamental matrix. Issue 6 (1st December 2014)

• Record Type: Journal Article
• Authors: Basah, Shafriza Nisha; Bab‐Hadiashar, Alireza; Hoseinnezhad, Reza
• Is Part Of: IET computer vision. Volume 8:Issue 6(2014)
• Page Start: 658
• Page End: 669
• Publication Date: 2014-12-01
• Subjects: motion estimation -- image segmentation -- computer vision -- Monte Carlo methods -- cameras -- matrix algebra; planar-motion segmentation analysis -- affine fundamental matrix -- computer vision -- multiple motion estimation -- dynamic scene images -- three-dimensional-rigid object motions -- camera parameters -- Monte Carlo simulation -- synthetic images -- Monte Carlo statistical method -- camera optical-axis -- general motion model; Computer vision -- Periodicals; Pattern recognition systems -- Periodicals; 006.37
• DOI: 10.1049/iet-cvi.2013.0224
• Languages: English
• ISSNs: 1751-9632

Abstract

Abstract : Various computer‐vision applications involve estimation of multiple motions from images of dynamic scenes. The exact nature of 3D‐object motions and the camera parameters are often not known a priori and therefore, the most general motion model (fundamental matrix) is applied. Although the estimation of fundamental matrix and its use for motion segmentation are established, the conditions for segmentation of different types of motions are largely unaddressed. In this study, we analysed the feasibility of motion segmentation using affine‐fundamental matrix, focusing on a scene includes multiple planar‐motions, viewed by an uncalibrated camera. We show that the successful segmentation of planar motion depends on several scene and motion parameters. Conditions to guarantee successful segmentation are proposed via extensive experiments using synthetic images. Experiments using real‐image data were set up to examine the relevance of those conditions to the scenarios in real applications. The experimental results demonstrate the capability of the proposed conditions to correctly predict the outcome of several segmentation scenarios and show the relevance of those conditions in real applications. In practice, the success of motion segmentation could be predicted from obtainable scene and motion parameters. Therefore these conditions serve as a guideline for practitioners in designing motion‐segmentation solutions.