A Gaussian mixture-based colour texture model

0106294 - UTIA-B 20040106 RIV US eng C - Conference Paper (international conference)
Haindl, Michal - Grim, Jiří - Somol, Petr - Pudil, Pavel - Kudo, M.
A Gaussian mixture-based colour texture model.
[Texturní model založený na gaussovských směsích.]
Proceedings of the 17th IAPR International Conference on Pattern Recognition. Vol. 3. Los Alamitos: IEEE, 2004 - (Kittler, J.; Petrou, M.; Nixon, M.), s. 177-180. ISBN 0-7695-2128-2.
[International Conference on Pattern Recognition /17./. Cambridge (GB), 23.08.2004-26.08.2004]
R&D Projects: GA ČR GA402/03/1310
EU Projects: European Commission(XE) 507752 - MUSCLE
Grant - others:Commission EC(XE) IST-2001-34744
Institutional research plan: CEZ:AV0Z1075907
Keywords : texture synthesis * mixture models
Subject RIV: BD - Theory of Information

A new method of colour texture modelling based on Gaussian distribution mixtures is discussed. We estimate the local statistical properties of the monospectral version of the target texture in the form of a Gaussian mixture of product components. The synthesized texture is obtained by means of a step-wise prediction of the texture image. The proposed texture modelling method can be viewed as a statistically controlled sampling.

Práce popisuje novou metodu pro modelování barevných textur založenou na pravděpodobnostním modelu typu gaussovské distribuční směsi. Odhadujeme lokální statistické vlastnosti monospektrální verze cílové textury ve formě gaussovské směsi součinových složek. Syntéza textury je založena na predikci texturního obrazu
Permanent Link: http://hdl.handle.net/11104/0013476