Modeling the pattern spectrum as a Markov process and its use for efficient shape classification

In this work the most important morphological granulometry, i.e. the pattern spectrum, is modeled, for the first time in the literature, as a first order Markov process. In addition, each of the terms of the process is shown to be normally distributed. The classification procedure followed for this specific application is based on modeling each separate class as a Markov process and making extensive use of the chain rule. Experimental results support the proposed classification procedure as quite promising, especially when compared to conventional classification techniques.