Seals contain a lot of vital information about the document. Seal detection is a necessary step to gain access to that information. In the present paper, we present a correlation based technique which exploits the existence of some constant character strings and their topology for detection of seals. We also present a technique which separates the false positives from the actual seals. We achieved an accuracy of about 98% for extraction of such kind of seals. Once the seal is extracted from a page image, an OCR has been employed to read the contextual information present in the seal. A knowledge based post-processing step is employed to enhance the accuracy of the recognized text strings.