IAM   02674
INSTITUTO ARGENTINO DE MATEMATICA ALBERTO CALDERON
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
TEMPORAL MODELING OF MOTION TEXTURES WITH MIXED-SATES MARKOV CHAINS
Autor/es:
TOMÁS CRIVELLI; BRUNO CERNUSCHI FRÍAS; PATRICK BOUTHEMY; JIAN-FENG YAO
Lugar:
Las Vegas, Nevada, USA, Abril 2008.
Reunión:
Congreso; IEEE. International Conference on Acoustics, Speech and Signal Processing (ICASSP'08), Pages 881-884; 2008
Institución organizadora:
IEEE
Resumen:
Dynamic textures are time-varying visual patterns that exhibitcertain spatio-temporal stationarity properties and aredisplayed mostly by natural scene elements. In this paper, wepresent new statistical models for the characterization of motionin this type of sequences. First we observe that motionmeasurements present values of two types: a discrete componentat zero expressing the absence of motion and a continuousdistribution for the rest of the motion values. Thus,we define random variables with mixed-states and propose tomodel a sequence of motion maps as a Markov chain, wherethe transition densities are mixed-state probability densities.Based on this approach, we propose a method for dynamictexture segmentation in real sequences showing the efficiencyof the proposal in dynamic content analysis applications.