“Segmentation of Motion Textures Using Mixed-State Markov Random Fields”

TOMAS CRIVELLI; BRUNO CERNUSCHI-FRÍAS; PATRICK BOUTHEMY; JIAN-FENG YAO

San Diego, California, USA

Congreso; SPIE "Conference on Mathematics of Data/Image Pattern Recognition, Compression, and Encryption with Applications", International Society for Optical Engineering, SPIE, San Diego, California, USA,, 17 al 25 de Agosto 2006. Paper No. 6315-23.; 2006

SPIE, International Society for Optical Engineering

The aim of this work is to model the apparent motion in image sequences depicting natural   dynamic scenes  (rivers, sea-waves, smoke, fire, grass etc) where some sort of stationarity and homogeneity of   motion is present.  We adopt the mixed-state Markov Random Fields models recently introduced to represent   so-called motion  textures. The approach consists in describing the distribution of some motion measurements   which exhibit  a mixed nature: a discrete component related to absence of motion and a continuous part for   measurements  different from zero.  We propose several extensions on the spatial schemes. In this context, Gibbs   distributions  are analyzed, and a deep study of the associated partition functions is addressed.   Our approach is valid for  general Gibbs distributions. Some particular cases of interest for motion texture modeling are   analyzed. This  is crucial for problems of segmentation, detection and classification.   Then, we propose an original approach for  image motion segmentation based on these models, where normalization factors are properly   handled. Results  for motion textures on real natural sequences demonstrate the accuracy and efficiency of our   method.  Nota: Versión extendida y con nuevos resultados de los trabajos presentados en AST'06 e ICIP'06.