Noisy word recognition using denoising and moment matrix discriminants

We consider the problem of recognition of a printed word belonging to a limited dictionary. The main difficulty comes from the fact that this word can be printed using different fonts, sizes, and positions on the page. Invariant moment methods for word recognition developed by Hu (1961) and Li (1992) are unreliable when the quality of the word image is degraded by noise. In this paper we investigate the effectiveness of simple median filter denoising for preprocessing noise degraded images prior to moment based classification using the moment matrix discriminants introduced by Hero et al. (1997)