离焦模糊图像中汉字识别技术研究

Study on the recognition technique of Chinese characters in out of focus image

提出了辐射投影变换(RPT),给出了数学模型和实现方法以及相应的投影函数(RPTF)。为了进一步提高其性能,对原始投影变换进行了优化改进,即累加原始函数的值并作旋转变换,从而得到了改进投影累加函数(IPAF),它可直接用于离焦模糊图像中汉字的识别。将傅里叶变换应用到IPAF上,使描述和识别用的特征量大为减少。采用基于IPAF的傅里叶子(前8个傅里叶系数)作为特征量,对离焦模糊的一级汉字进行识别,总识别率达到了97 4%。

An algorithm called Radiant projection transformation (RPT) is proposed, and the mathematical model and implementation method, as well as the radiant projection transformation function (RPTF) are given, which can be used in the recognition of Chinese character in out of focus image. The performance of the original algorithm is improved and optimized by accumulating the value of RPTF and applying a rotation transform. Consequently a new function called improved projection accumulative function (IPAF) is obtained. Fourier transform is applied on IPAF, and as a result, the amount of data for describing and recognizing shapes is reduced vastly. Based on theoretical analysis and experiments, Fourier descriptors based on IPAF are utilized as features for recognition. The Fourier descriptors accuracy used Chinese characters (out of focus) reaches 974% by using Fourier descriptors based on IPAF.