Fast Learning of Sprites using Invariant

Moray Allan; Michalis K Titsias; Christopher K I Williams

Fast Learning of Sprites using Invariant

Moray Allan, Michalis K Titsias, Christopher K I Williams

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

A popular framework for the interpretation of image sequences is the layers or sprite model of e.g. Wang and Adelson (1994), Irani et al. (1994). Jojic and Frey (2001) provide a generative probabilistic model framework for this task, but their algorithm is slow as it needs to search over discretized transformations (e.g. translations, or affines) for each layer. In this paper we show that by using invariant features (e.g. Lowe’s SIFT features) and clustering their motions we can reduce or eliminate the search and thus learn the sprites much faster. We demonstrate our algorithm on two image sequences.

Original language	English
Title of host publication	Proceedings of the British Machine Vision Conference 2005
Publisher	BMVA Press
Number of pages	10
ISBN (Print)	1 901725 29 4
Publication status	Published - 2005
Event	British Machine Vision Conference 2005 (BMVC) - Oxford Brookes Univeristy, Oxford, United Kingdom Duration: 5 Sept 2005 → 8 Sept 2005

Conference

Conference	British Machine Vision Conference 2005 (BMVC)
Country/Territory	United Kingdom
City	Oxford
Period	5/09/05 → 8/09/05

Access to Document

paper247Accepted author manuscript, 391 KB

http://www.robots.ox.ac.uk/~phst/BMVC2005/papers/paper-57-247.html

Cite this

@inproceedings{4db335d433f54617832f75181f73b355,

title = "Fast Learning of Sprites using Invariant",

abstract = "A popular framework for the interpretation of image sequences is the layers or sprite model of e.g. Wang and Adelson (1994), Irani et al. (1994). Jojic and Frey (2001) provide a generative probabilistic model framework for this task, but their algorithm is slow as it needs to search over discretized transformations (e.g. translations, or affines) for each layer. In this paper we show that by using invariant features (e.g. Lowe{\textquoteright}s SIFT features) and clustering their motions we can reduce or eliminate the search and thus learn the sprites much faster. We demonstrate our algorithm on two image sequences.",

author = "Moray Allan and Titsias, {Michalis K} and Williams, {Christopher K I}",

year = "2005",

language = "English",

isbn = "1 901725 29 4 ",

booktitle = "Proceedings of the British Machine Vision Conference 2005",

publisher = "BMVA Press",

note = "British Machine Vision Conference 2005 (BMVC) ; Conference date: 05-09-2005 Through 08-09-2005",

}

TY - GEN

T1 - Fast Learning of Sprites using Invariant

AU - Allan, Moray

AU - Titsias, Michalis K

AU - Williams, Christopher K I

PY - 2005

Y1 - 2005

N2 - A popular framework for the interpretation of image sequences is the layers or sprite model of e.g. Wang and Adelson (1994), Irani et al. (1994). Jojic and Frey (2001) provide a generative probabilistic model framework for this task, but their algorithm is slow as it needs to search over discretized transformations (e.g. translations, or affines) for each layer. In this paper we show that by using invariant features (e.g. Lowe’s SIFT features) and clustering their motions we can reduce or eliminate the search and thus learn the sprites much faster. We demonstrate our algorithm on two image sequences.

AB - A popular framework for the interpretation of image sequences is the layers or sprite model of e.g. Wang and Adelson (1994), Irani et al. (1994). Jojic and Frey (2001) provide a generative probabilistic model framework for this task, but their algorithm is slow as it needs to search over discretized transformations (e.g. translations, or affines) for each layer. In this paper we show that by using invariant features (e.g. Lowe’s SIFT features) and clustering their motions we can reduce or eliminate the search and thus learn the sprites much faster. We demonstrate our algorithm on two image sequences.

M3 - Conference contribution

SN - 1 901725 29 4

BT - Proceedings of the British Machine Vision Conference 2005

PB - BMVA Press

T2 - British Machine Vision Conference 2005 (BMVC)

Y2 - 5 September 2005 through 8 September 2005

ER -

Fast Learning of Sprites using Invariant

Abstract

Conference

Access to Document

Fingerprint

Cite this