Self Acquiring Image Knowledge Using Maximum Value Metric based Self Organizing Maps

Mining images means extracting patterns and derive knowledge from large collections of images. Image mining follows image feature gathering, learning and retrieving procedures. This paper apprises as to what extent the users of the self organizing maps(SOM) techniques are satisfied with its efficiency of visualizing and organizing large amounts of image data. The main contribution of the paper consists of identifying factor that influences the quality of SOM The result analysis shows that, SOM learning capacity is sensitive to initial weight vector, Learning rate, number of epochs for training and distance measure to select winning neuron. The result affirms that among all theses features distance measure factor has high rate of impact in SOM clustering. Euclidian measure is substituted by the Linfin norm (maximum value distance) measure of Minkowski r_metric. Maximum value distance based SOM exhibits both accurate functionality and image mining feasibility.