Title :
Unsupervised texture segmentation using stochastic version of the EM algorithm and data fusion
Author :
Cruz, Carlos Avils
Author_Institution :
Dept. de Electr., Univ. Autonoma Metropolitana, San Pablo, Mexico
Abstract :
In this paper I present a new methodology for texture segmentation. This methodology is based through the high order statistics features, the data fusion techniques and finally though the maximum likelihood method in order to find the clusters. The methodology is applied in order to segment natural micro-textures
Keywords :
higher order statistics; image segmentation; image texture; maximum likelihood estimation; sensor fusion; clusters; data fusion; expectation maximisation; high-order statistics features; maximum likelihood method; natural micro-texture segmentation; stochastic EM algorithm; unsupervised texture segmentation; Character recognition; Delay estimation; Feature extraction; Frequency domain analysis; Parallel architectures; Radar; Robots; Robustness; Statistics; Stochastic processes;
Conference_Titel :
Pattern Recognition, 1998. Proceedings. Fourteenth International Conference on
Conference_Location :
Brisbane, Qld.
Print_ISBN :
0-8186-8512-3
DOI :
10.1109/ICPR.1998.711859
