Noise suppression over bi-level graphical documents using a sparse representation

In this paper, we explore the use of learning algorithm (K-SVD) for building dictio- naries adapted to the image properties. In addition, in our model, we also modeled the energy of the noise basing on the function of the normalized cross-correlation between noised and non noised documents identified in training set. We have evaluated this method on the Grec2005 dataset. The experimental results demonstrate the robustness of our approach by comparing it with state-of-the-art methods.