An Efficient Method for Urdu Language Text Search in Image Based Urdu Text

This paper describes an efficient method for Urdu text search in computer generated and handwritten scanned images. An efficient text search technology is necessary because of increasing handled document every day. This method is unique and simple in the sense that no features are extracted. The proposed method is script independent. The input image is directly matched with a set of prototype characters representing each possible class. The distance between each input image and each prototype character is computed, and the character is assigned to the class of the prototype giving the best match. Experimental results show 100 % accuracy for 4, 5-character ligatures, 87 % for 3-character ligature and 78 % for 2-character ligatures.