Image segmentation via X-means under overlapping classes

摘要

To provide tools for image understanding, non-trivial task of image segmentation is now put on a new semantic level of object detection. Internal, external and contextual region properties often can adequately represent image content but there arises field of view coverings. Truthful image interpretation strictly depends on valid number of regions. The goal is an attempt to solve image clustering problem under fuzzy conditions of overlapping classes.