Character segmentation plays a very important role in a text recognition system. The simple technique of using inter-character gap for segmentation is useful for fine printed documents, but this technique fails to give satisfactory results if the input text contains touching characters. In this paper, we have proposed two algorithms to segment touching characters, and one algorithm to segment overlapping lines in degraded printed Gurmukhi document. Various categories of touching characters in different zones, along with their solutions, have been proposed. The solution methodology extensively uses the structural properties of Gurmukhi script. The algorithm proposed for segmenting horizontally overlapping lines uses a heuristics based upon the height of a character. The problem of multiple horizontally overlapping lines may occur in a number of situations such as printed newspapers, old magazines and books etc. Similarity among Indian scripts allows us to use these algorithms for solving the segmentation problems in other Indian languages also.
Real-time object detection and face recognition system to assist the visually impaired
