摘要： 提出了一种基于投影归一化的字符特征提取方法，该方法首先对字符图像进行横向扫描和纵向扫描生成行投影向量和列投影向量，然后通过对行投影向量和列投影向量进行维数和密度的归一化处理生成双投影归一化向量作为特征向量。聚类和识别实验表明双投影归一化向量不仅计算简单，而且对同种字体不同字号的英文字符识别可达到较好的结果。

关键词: 特征提取；行投影向量；列投影向量；归一化；双投影归一化向量

Abstract: This paper presents a character feature extraction method based on projection normalization. In this method, a character image is scanned in both horizontal and vertical directions to create a row-projection vector and a column-projection vector which are further normalized in dimension and density to create a double projection normalized vector, namely, the feature vector. Some experiments show that double projection normalization vectors are not only easy to be calculated, but also good for recognizing English characters with the same font but in different sizes

Key words: Feature extraction; Row-projection vectors; Column projection vectors; Normalization; Double projection normalization vectors