Word Spotting in Scanned Images of Printed Devanagri Documents
Premendra Tiwari, Siddharth H. Patel, Sunny Sangwan · 2015
In this paper, we propose a keyword retrieval system for locating words in printed Devanagari document images. Government has initiated various schemes to scan and digitally preserve such documents for future use. Such scanned images of documents are now available to users. Even though the documents are available in their digital format, it is still difficult to search for a single word or phrase as they are scanned images. Traditional optical character recognition techniques (OCR) and other text retrieval systems fail on such type of document images due to various types of noises. In such situations word spotting shall be a major help to users to automatically search for a particular word/phrase in millions of such document images. An attempt is made here to design and implement a word spotting technique for printed Devanagari documents. Based on the word spotting technology, a collection of document images is converted into a collection of word images by word segmentation, and a number of profilebased features are extracted to represent word images. Dynamic Time Warping(DTW) is applied for image comparisons. Imageto-image matching is done by calculating similarities between a query word image and each word image in the collection, and consequently, a ranked result is returned in descending order of the similarities. The system supports GUI. The optimal performance of the system is validated.