Clustering techniques provide an unsupervised approach for extracting information from images in color space representation, which is useful in applications including image analysis, quantization, and compression. We present our algorithm for parallel computation of a regularized k-means, which takes advantage of modern computing hardware to efficiently produce clusters while adaptively choosing the number of clusters which is not known a priori. This is advantageous when the optimal partitioning of the space is not known. We discuss the properties of clusters produced by the algorithm and examples drawn from imaging applications.

This presentation is part of Minisymposium “MS3 - Applications of Imaging Modalities beyond the Visible Spectrum (2 parts)”
organized by: Max Gunzburger (Florida State University) , G-Michael Tesfaye (Naval Surface Warfare Center, Panama City) , Janet Peterson (Florida State University) .

Authors:

Benjamin McLaughlin (NSWC-PCD)

Keywords:

image quantization, parallel computing, unsupervised learning