Abstract

In the field of image processing, invariance refers to the properties of an image that remain unchanged or exhibit minimal differences under transformations like rotation, scaling, or blurring. The primary focus of this work is to overcome the challenge of recognizing handwritten characters from various rotation angles. To address this, we propose a novel approach that utilizes modified quad-vector features proposed in [7], enabling the accurate identification of pixels associated with specific rings. Our main objective is to achieve efficiency and the rotation invariance too. To accomplish this, we employ only two vectors to extract features rather than four as suggested in [7]. Hence the name dual vector features are given. The extracted dual vector features establish a framework for extracting rotation-invariant features (RIFS) from black-and-white images. The proposed approach is found 1.13 times faster than the original method [7] and thus efficient and recommended for Handwritten Character Recognition (HCR) systems