Machine learning and adaptive morphological operators

Abstract

This work proposes the use of machine learning methods applied to the construction of a gray level adaptive hit-or-miss morphological operator. Because they are adaptive and translation invariant, it is expected that these operators can be better utilized for the process of pattern recognition. In a first approach, the investigated adaptive model is inspired on the Vector Quantization Unsupervised Learn Rule and developed through Elementary Look-Up Tables (ELUTs) formalism of elementary morphological operators in gray level images.