Date Added: Apr 2012
The aim of this paper is to present a new approach for modal regions detection of histograms representing multidimensional data distributions. Based on the Markov field theory, this approach considers the histograms as a field Y of measures. The hidden field X is used to model the modal regions and the valleys of the histogram. The Iterated Conditional Modes algorithm (ICM), combined with the Estimation-Maximization algorithm (EM), makes it possible to detect the local maxima of the histogram and allows determining the identified hidden field X from Y, which is composed of the modal regions and the valleys of the histogram.