Morphological texture-based maximum-likelihood pixel classification based on local granulometric moments

摘要: Abstract Morphological granulometries are one-parameter filter sequences that monotonically decrease image area. A size distribution is generated by measuring the residual area after each iteration of sequence. Normalization yields a probability function whose moments can be employed as signatures. By locally over window about point an instead entire image, local texture features at pixel, and these for pixel classification. using several granulometric structuring-element generating sequences, numerous moment sets result, carrying different textural information. detailed analysis this classification methodology Gaussian maximum likelihood classifier provided. Included statistical study accuracy, feature optimization, robustness with respect to various relevant noise models.