Authors:
Lalit Kumar
;
Abhay Bansal
;
Dinesh Ganotra
and
Neeraj Jain
Affiliation:
Newgen Software Technologies Limited, India
Keyword(s):
Binarization, Thinning, Skeleton, Local Coupling Point, LCP, ICR, OCR, Offline Signature Verification.
Abstract:
Image thinning, while preserving geometrical properties of the image, remains one of the most challenging areas in Image Processing. In this paper, we introduce an image thinning approach that takes as input gray scale images and produces binary thinned images as output. Our approach uses bi-directional iterative thinning process to get single-pixel-thinned image (primary skeleton), while preserving the geometric properties of the image. The gray scale image is binarized using dyamic threshold, and further processed to remove noise. Black pixels are classified as contour pixels (pixels exposed to the background) and body pixels (black pixels other than the contour pixels). Thinning process involves scanning contour pixels from two opposite directions simultaneously, while preserving Local Coupling Points (LCP) and removing the rest of pixels.