Slant Correction Of Gurmukhi Text Words From Natural Scene Images
Singh Balwinder, MAINI RAMAN · i-manager s Journal on Information Technology · 2017
Slant correction is generally performed to normalise handwritten characters to improve the results of recognition. But, natural scene images containing printed text suffer from both skew and slant deformation due to 3D tilt and projective transformation. In this study, a slant correction technique for natural scene images containing Gurmukhi text words has been proposed.The skew normalized Gurmukhi words are firstly transformed to salient image and Hough transform is then applied onto edge image obtained through Sobel operator in vertical direction. Of all the lines identified by Hough transform, the angle of longest line with vertical axis is inferred as the slant angle of Gurmukhi word. Horizontal affine shear transformation with previously determined slant angle is carried out to correct slant in word image. Experimental results show that the method is very effective for slant correction of Gurmukhi words from scene images. The method works equally well for other headline based Indian scripts like Devanagari as well without any modification.