Gianfranco Doretto / Publications

Dynamic texture modeling

Doretto, G.
Dynamic texture modeling
Master's Thesis, University of California, Los Angeles, CA,2002. Committee: Adnan Darwiche, Michael Dyer, Stefano Soatto (Chair).

Download

PDF (6.8MB )  

Abstract

Dynamic textures are sequences of images of moving scenes that exhibit certain stationarity properties in time; these include, for example, sea-waves, smoke, foliage, whirlwind etc. This work presents a novel characterization of dynamic textures that poses the problems of modeling, learning, recognizing, classifying, synthesizing, and editing dynamic textures on a firm analytical footing. By means of system identification tools it is possible to capture the “essence” of dynamic textures; this is done by learning (i.e. identifying) models that are optimal in the sense of maximum-likelihood or minimum prediction error variance. For the special case of second-order stationary processes, a model can be identified sub-optimally in closed-form. Once learned, a model has predictive power and can be used for extrapolating, and editing (i.e. modifying the temporal and spatial behavior of) synthetic sequences. It is presented experimental evidence that, within this framework, even low-dimensional models can capture very complex visual phenomena. Furthermore, it is shown the possibility to map the manipulation of model parameters into sensible changes of visual appearance in extrapolated sequences. The uniqueness of the model allows to pose the problem of recognition and classification in the space of models. Since the space is non-linear, a distance between models must be defined. This work examines three different distances in the space of autoregressive models and assess their power.

BibTeX

@MastersThesis{doretto02thesis,
  Title                    = {Dynamic texture modeling},
  Author                   = {Doretto, G.},
  School                   = {University of California},
  Year                     = {2002},
  Address                  = {Los Angeles, CA},
  Month                    = {June},
  Note                     = {{C}ommittee: {A}dnan {D}arwiche, {M}ichael {D}yer, {S}tefano {S}oatto ({C}hair).},
  Abstract                 = {Dynamic textures are sequences of images of moving scenes that exhibit certain stationarity properties in time; these include, for example, sea-waves, smoke, foliage, whirlwind etc. This work presents a novel characterization of dynamic textures that poses the problems of modeling, learning, recognizing, classifying, synthesizing, and editing dynamic textures on a firm analytical footing. By means of system identification tools it is possible to capture the “essence” of dynamic textures; this is done by learning (i.e. identifying) models that are optimal in the sense of maximum-likelihood or minimum prediction error variance. For the special case of second-order stationary processes, a model can be identified sub-optimally in closed-form. Once learned, a model has predictive power and can be used for extrapolating, and editing (i.e. modifying the temporal and spatial behavior of) synthetic sequences. It is presented experimental evidence that, within this framework, even low-dimensional models can capture very complex visual phenomena. Furthermore, it is shown the possibility to map the manipulation of model parameters into sensible changes of visual appearance in extrapolated sequences. The uniqueness of the model allows to pose the problem of recognition and classification in the space of models. Since the space is non-linear, a distance between models must be defined. This work examines three different distances in the space of autoregressive models and assess their power.},
  Bib2html_pubtype         = {Theses},
  Bib2html_rescat          = {Dynamic Textures, Visual Motion Analysis, Image Based Rendering},
  File                     = {doretto02thesis.pdf:doretto\\thesis\\doretto02thesis.pdf:PDF;doretto02thesis.pdf:doretto\\thesis\\doretto02thesis.pdf:PDF},
  Owner                    = {doretto},
  Timestamp                = {2007.01.19}
}