Arabic documents processing is one of the urgent tasks to be reviewed today. Spread of digitized resources makes the informational society use modern technologies in order to retrieve exact data within a minimum period. Documents processing demands developing of new approaches on the bases of analysis of the Arabic graphic in particular and peculiarities of the Arabic language as a whole. Currently different principles of document processing in Arabic are applied. Some of them are based on the grammar structure of the language, while others use techniques that rely on images classification. The paper critically analyses main techniques for different types of Arabic document classification and suggests the most efficient methods for their processing. Choice of the most relevant approach depends on the character of a certain extra linguistic task, for example principles of classification of modern handwritten documents varies from the technique, used for historical manuscript processing.