The talk presents numerical methods and tools for an automatic Middle-Age handwriting processing: from the contrast enhancement to the writing styles classification. The challenges consist in proposing solutions for an efficient paleographical handwritings classification that embeds both robust shapes description, consistent similarity measures and ability to produce relevant results on a great variety of styles. Constraints of knowledge quantity (only based on the writing dynamics study) and images quality (degraded documents) have guided our researches on different directions that are described and compared.