Adversarial Autoencoders for Denoising Digitized Historical Documents: The Use Case of Incunabula
Hala Neji, Javier Nogueras-Iso, Javier Lacasta, Mohamed Ben Halima, Adel M. Alimi · 2019
Historical document denoising is the most challenging step in the field of image processing and computer vision. In this paper, we propose a novel end-to-end adversarial autoencoder (AAE) to generate clean images and to show how adversarial autoencoders can be used in historical document denoising. We used the Adversarial Autoencoder (AAE), which uses the generative adversarial networks (GAN) so as to suit the aggregated posterior of the hidden code vector of the autoencoder with an arbitrary prior. The experiments results prove that our approach functions more positively than the cutting-edge approaches on synthetic and real world images at a lower computational cost as well.