Trademark identification through text recognition

Abstract

Trademark identification is of great interest in the document domain due to two reasons. Given a document that consists of a trademark, the trademark needs to be conclude whether present or not in the database and given a known trademark, we need to index into a database of documents and extract all information related to it. This paper proposes a new method to recognise the text on mixed-mode trademarks that consists of various font style and sizes, focusing on uppercase Roman alphabets. A stroke-based feature extraction method represented by chain codes is used. A modified rule-based classifier recognises the characters with satisfactory results.