Touching text character localization in graphical documents using SIFT

摘要: Interpretation of graphical document images is a challenging task as it requires proper understanding text/graphics symbols present in such documents. Difficulties arise recognition when text and symbol overlapped/touched. Intersection with lines curves occur frequently documents hence separation very difficult. Several pattern classification techniques exist to recognize isolated text/symbol. But, the touching/overlapping has not yet been dealt successfully. An interesting technique, Scale Invariant Feature Transform (SIFT), originally devised for object can take care overlapping problems. Even if SIFT features have emerged powerful descriptors, their employment context investigated much. In this paper we adaptation approach character localization (spotting) We evaluate applicability technique discuss scope improvement by combining some state-of-the-art approaches.