A Cluster Grouping Technique for Texture Segmentation

We propose an algorithm for texture segmentation based on a divide-and-conquerstrategy of statistical modeling.Selectedsets of Gaussianclusters,estimated via ExpectationMaximization on thetexture features, aregrouped together to form compositetextureclasses.Ourclustergroupingtechniqueexploits the inherent localspatialcorrelationamongposterior distributions of clusters belonging to the same texture class.Despiteitssimplicity,thisalgorithm canmodel even very complex distributions, typicalof natural outdoor images.

• Type: Conference
• Year: 2000
• Where: ICPR
• Authors: Roberto Manduchi
• Keywords: Complex Distributions, Computer Vision, ICPR 2000, Inherent Localspatialcorrelationamongposterior Distributions, Texture Segmentation