Adaptive normalization of handwritten characters using global/local affinetransformation

Conventional normalization methods for handwritten characters have limitati ons as preprocessing operations because they are category-independent. This paper introduces an adaptive or category-dependent normalization method th at normalizes an input pattern against each reference pattern using global/ local affine transformation (GAT/LAT) in a hierarchical manner as a general deformation model. Also, the normalization criterion is clearly defined as minimization of the mean of nearest-neighbor interpoint distances between each reference pattern and a normalized input pattern. According to the abo ve-mentioned criterion, optimal GAT/LAT is determined by iterative applicat ion of weighted least-squares fitting techniques. Experiments using input p atterns of 3,171 character categories, including Kanji, Kana, and alphanume rics, written by 36 people in the cursive style against square-style refere nce patterns show not only that the proposed method can absorb a fairly lar ge amount of handwriting fluctuation within the same category, but also tha t discrimination ability is greatly improved by the suppression of excessiv e normalization against similarly shaped but different categories. Furtherm ore, comparative results obtained by the conventional shape normalization m ethod for preprocessing are presented to show the superiority of the propos ed category-dependent GAT/LAT normalization over category-independent norma lization.