Motion Texture Segmentation with Simultaneous Parameter Estimation

AGUSTÍN MAILING; BRUNO CERNUSCHI FRÍAS

Río Gallegos, Provincia de Santa Cruz, Argentina

Congreso; RPIC 2007, XII Reunión de Trabajo en Procesamiento de la Información y Control, 16 al 18 de octubre de 2007; 2007

Universidad Nacional de la Patagonia Austral

In this work a method for mixed-state  model motion texture segmentantion and parameter  estimation is presented. We use the Expectation Maximization  algorithm for mixture parameter estimation  introducing the Gibbs distribution for moving  points, excluding zero discrete component associated  with no motion regions. We use then the a posteriori  probabilities to generate an alternative field to segment  the textures according to its statistical parameters.