We propose an approach dedicated to recognize characters from binary images by an hybrid strategy. A statistical method is used to identify the global shape of each alphanumeric symbol. This approach is based on the Fourier Descriptor computation of the outer and potentially inner contours extracted from the input image. %The mapping to the Fourier space enables us to smooth the signal of the character boundary and makes the approach robust to noise. The recognition is managed by a Hierarchical Neural Network (HNN), that is able to deal with topological errors in the contour extraction. This strategy is extremely efficient for the majority of the classes: the recognition rate reaches about 99.5%. However, the performances sensitively decrease for 'similar characters', i.e 8/B. For these classes, the statistical strategy is not adapted for performing an accurate recognition. Indeed, not only the binary images are hard to distinguish, but the Fourier Descriptor extraction emphasizes the difficulty to discriminate sailent features such as corners. For these characters, we adopt a strategy that revolves around decomposing the characters into structural elements. The Reeb graph generated from the binary images and a simple polygonal approximation permit to capture both topological and geometrical relevant features. The classification stage is carried out by a boosting algorithm (ADABOOST). Several results validate the conjoint use of the structural/statistical approach for character recognition, and point out the relevance of the strategy for a general pattern recognition purpose.