The MorphoNet project is an internal project of the Image Processing and Pattern Recognition group, aiming to bridge the gap between classical morphological operations (such as erosion and dilation, as well as their compositions like opening and closing) and neural networks. The goal of the project is to develop trainable morphological operators (in the spirit of classical convolution layers in convolutional neural networks), with the ultimate objective of enabling the automatic learning of sequences of morphological operations.

Structure of a morphological neural network with morphological layers designed to learn morphological operations