Unsupervised image segmentation using the modified pyramidal linking approach

Abstract

As image segmentation is one of the first steps in image processing, little knowledge about the image content is available. The modified pyramidal linking approach presented here is intended as a low-level process for segmentation without any a priori information or user-defined thresholds. The process works with multiple features which are weighted according to their information content. The number of extracted regions is automatically determined by the process itself.