Unsupervised Classification of X-Ray Mapping Images of Polished Sections

X-ray mapping images of polished sections are classified using two unsupervised clustering algorithms. The methods applied are the k-means algorithm and an extended spectral fuzzy c-means algorithm. The extentions include new types of memberships that are related to the contextual information. In addition to the traditional spectral membership we apply a spatial membership and a parental membership. The parental membership is introduced by implementing the algorithm in a scale-space representation. Both spectral and spatial information is carried across levels, enhancing speed and visual impression of the segmentation of image data.

• Type: Conference
• Year: 2000
• Where: MVA
• Authors: Klaus Baggesen Hilger, Allan Aasbjerg Nielsen, Jens Michael Carstensen
• Keywords: MVA 2000, MVA 2007, Parental Membership, Spectral Fuzzy C-means, Traditional Spectral Membership