A fuzzy approach to segment document images

A new method to segment characters and numbers in the presence of a coloured background, texture and noise is presented. It also shows that it is able to work correctly on a document with overlapped static and dynamic structures. This method, based on the concept of fuzzy connectivity exploits both intensity information and continuity of direction which characterise structures belonging to the document. An image-scanning mechanism that is non-iterative and non-causal is used to propagate the information by following paths with maximum connectivity. The final result of this process is a grey level image ( the "connectedness" image) where every value represents the degree of membership of the pixel to the object searched.