An Introduction to Deep Morphological Networks

Over the past decade, Convolutional Networks (ConvNets) have renewed perspectives of research and industrial communities. Although this deep learning technique may be composed multiple layers, its core operation is convolution, an important linear filtering process. Easy fast to implement, convolutions actually play a major role, not only in ConvNets, but digital image processing analysis as whole, being effective for several tasks. However, aside from convolutions, researchers also proposed developed non-linear filters, such operators provided by mathematical morphology. Even though these are so computationally efficient general, they able capture different patterns tackle distinct problems when compared convolutions. In paper, we propose new paradigm networks where replaced morphological filters. Aside performing operation, Deep Morphological Network (DeepMorphNet) learn filters (and consequently features) based on input data. While process raises challenging issues regarding training actual implementation, DeepMorphNet proves extract features solve that traditional architectures with standard convolution cannot.