Optimized Gurmukhi Text Recognition from Signboard Images Captured by Mobile Camera Using Structural Features

Earlier, research was restricted to the images acquired by traditional scanners, however an innovative trend of research has emerged with the evolution of portable, high speed digital cameras and multimedia mobile phones comprising smart features. They provided us the opportunity to employ them for image acquisition as an alternate to traditional scanners for the recognition purpose. This subject has attracted numerous researchers, meanwhile it provides a means for automatic processing of substantial amount of data. Text to speech translation of recognized text from images can be ready to lend a hand for visually impaired people and for those who are unfamiliar with the language. This paper provides technical solution for the recognition of Gurmukhi text from the images of different signboards acquired by camera of different resolution. Segmentation is accomplished using vertical and horizontal projection histograms on the pre-processed image which breakdowns the text into lines, words and characters. Feature extraction and recognition on the segmented characters is accomplished by considering at least three corresponding structural features holes, endpoints and junctions. Consequently, our recognition is grounded on the location and number of these features extracted. The proposed algorithm was tested on 1300 images of Gurmukhi text acquired by camera and recognition rate of 90% demonstrates the precision of the system.