Generic Markov-Gibbs Model and its Identification

This chapter introduces a generic Markov-Gibbs random field (MGRF) model of spatially homogeneous or piecewise-homogeneous textures  [37,36], and its identification. Both the traditional model identification via MCMC algorithms and a novel structural identification are discussed. The new structural identification results in a texel-based texture description and a fast texture synthesis technique. Compared to the traditional model identification via MCMC algorithms, the proposed method is light-weighted and much faster in both model identification and texture synthesis. This work has been reported in [40,108,109,110,111]. A new and more precise model identification approach, developed recently^6.1, is briefly outlined in Appendix A.