Parallelizing a new environment based clustering method

There exists a wide range of problems which requires the automatic classification of a data set. In this sense, clustering techniques have been applied, since they are characterized by forming classes or groups using a predefined similarity measure. The present article presents algorithm architecture and structure for paralleling clustering algorithm EBC (environment based clustering) which, deferring from usual solutions, processes input patterns in order to establish the similarity measure to be used. Results obtained are analyzed over images of liver tissues with a maximum range of 256 colors, studying algorithm dependence on image resolutions and the number of different patterns in them. Then, critical points of the sequential algorithm are optimized over a PC net architecture. Finally, the extension of the results obtained are discussed, as well as the solution presented for the case of high resolution images, in which the number of different patterns is of higher order (between 3000 and 5000). Eje: Programación concurrente Red de Universidades con Carreras en Informática (RedUNCI)