Consensus Image Feature Extraction with Ordered Directionally Monotone Functions.

In this work we propose to use ordered directionally monotone functions to build an image feature extractor. Some theoretical aspects about directional monotonicity are studied to achieve our goal and a construction method for an image application is presented. Our proposal is compared to well-known methods in the literature as the gravitational method, the fuzzy morphology or the Canny method, and shows to be competitive. In order to improve the method presented, we propose a consensus feature extractor using combinations of the different methods. To this end we use ordered weighted averaging aggregation functions and obtain a new feature extractor that surpasses the results obtained by state-of-the-art methods.