Morphologically constrained GRFs: Applications to texture synthesis and analysis

作     者：Sivakumar, K Goutsias, J 

作者机构：Washington State Univ Sch Elect Engn & Comp Sci Pullman WA 99164 USA Johns Hopkins Univ Ctr Imaging Sci Baltimore MD 21218 USA Johns Hopkins Univ Dept Elect & Comp Engn Baltimore MD 21218 USA 

出 版 物：《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 (IEEE Trans Pattern Anal Mach Intell)

年 卷 期：1999年第21卷第2期

页      面：99-113页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：Office of Naval Research  ONR  (N00014-90-1345) 

主　　题：Gibbs random fields mathematical morphology Metropolis algorithm Monte Carlo simulation morphological constraints size density statistical inference texture synthesis and analysis 

摘      要：A new class of Gibbs random fields (GRFs) is proposed capable of modeling geometrical constraints in images by means of mathematical morphology. The proposed models, known as morphologically constrained GRFs, model images by means of their size density. Since the size density is a multiresolution statistical summary morphologically constrained GRFs explicitly incorporate multiresolution information into image modeling. Important properties are studied and their implication to texture synthesis and analysis is discussed. For morphologically constrained GRFs to be useful in practice, it is important that an efficient technique is available for fitting these models to real data. It is shown that, at low enough temperatures and under a natural condition, the maximum-likelihood estimator of the morphologically constrained GRF parameters can be approximated by means of an important tool of mathematical morphology known as the pattern spectrum. Therefore, statistical inference can he easily implemented by means of mathematical morphology. This allows the design of a, computationally simple morphological Bayes classifier which produces excellent results in classifying natural textures.