Unsupervised learning algorithms for comparison and analysis of images

This paper proposes a computational scheme for comparison and color analysis of images by using unsupervised learning algorithms. As a first step, two special growing unsupervised learning algorithms are introduced and used to create the so called compressed information model (CIM) which replaces the original ldquoraw datardquo (the RGB pixels) of the image with a much smaller number of neurons. Then two main features are extracted from the CIM, namely the center-of-gravity of the model and the weighted average size. It is shown in the paper that they can be used separately or in a combined way (in a fuzzy decision block) for a more precised similarity analysis between pairs of images. Another type of image analysis is also described in the paper that uses the unsupervised learning algorithm to generate preliminary fixed small number of neurons (regarded as key-points). They define the most important color areas in the RGB space which show important color details of the image. The whole proposed computational scheme in the paper is demonstrated on a test example consisting of 6 images of different flowers and trees.