Towards Automatic Fractal Feature Extraction for Image Recognition

Fractal encoding based on Iterated Function Systems is a well-known method which is currently exploited mainly for image compression tasks. The reason is that it allows to produce very compact encodings without any meaningful loss of precision. In this work, we present the results of a study aimed at showing that, within the field of image analysis, it is possible to use fractal encodings based on Iterated Function Systems also to extract information that is relevant in object recognition tasks. In particular, we propose a new kind of features, called fractal features, to be used jointly with machine learning techniques in isolated 2D image recognition tasks and show that it is possible to extract them in an automatic way.