7 - Approche spatio-temporelle pour l'analyse de séquences d'images. Application en détection de mouvement
Year: 1997
Type: article
Abstract: In this paper, a new strategy for motion analysis in image sequences is proposed . \nThe originality of this work is that an image sequence is seen as a 3D data flow instead of a series of 2D-images . By applying such \nan approach to the problem of motion detection in image sequences acquired with a static camera, a spatiotemporal Markovian \nmodel associated with spatiotemporal relaxation is defined . Thanks to an adequate modelling of spatiotemporal interactions \nbeetween pixels belonging to a 3D neighbourhood, good results may be obtained for detecting moving objects in noisy sequences \nor slow moving objects . \nIn order to improve the ability of the algorithm to detect poorly textured objects or subpixel moving objects, it may be integrated \nin a multiresolution scheme . The data pyramid is built by using 3D low-pass filters and 3D subsamplings . Results on synthetic and \nreal worl image sequences are exhibited, showing the validity of this approach .
Source: Traitement du signal
Authors: Caplier, Luthon
Language: French
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2244613777
