Achieving Premier Invariance to Scale and Rotation for Nandinagari Character Recognition by Comparing Multi Moment Features

Any superlative recognition system would need to achieve top tolerance to changes in Scale and Rotation of the image. In this paper we discuss handwritten Nandinagari manuscripts and their moment invariant features. The proposed work is first of its kind where Moment Invariant Features are applied to these rare characters, which were existing much before Devanagari characters. Here, we have used four different moments namely Hu Moments, Zernike Moments, Legendre Moments and Tchebichef Moments. For the purpose of study, images of the characters are subjected to different rotations and scaling. Then, moment invariant features are calculated. The process is repeated for all the moment invariant methods and critically examined. Analysis of results show that features obtained by applying Tchebichef moment technique remains invariant with different rotation and scaling as compared to other three moment techniques.