Detection of structures using a robust edge detection algorithm with expansion of border windows

Searching specific structures of different scales in digital images, we tested multiple alternatives of a mobile window method, developed starting from an algorithm by Cayula and Cornillon. These alternatives presented an increasing computing cost when detecting meso-structures in complex images. Quality borders required processing a great number of window-threshold pairs, because to detect some structures was required to relax the work conditions (bimodality and cohesion). Our proposal, in this work, is relaxing the conditions but not for all the window-threshold pairs, only to a selected set of windows, those near windows that accomplish strict conditions. We propose a new method that is a combination of the mobile window one, previously used, and the seeded regions approach. In this case, we seed not regions but borders.