摘要
提取稳定而有代表性的特征是视频图像字符识别的核心问题之一。文章提出了一种基于小波和矩的图像字符特征向量提取方法。通过对字符图像的不同小波分解子图求取不同的矩特征,构造出字符的特征向量。该方法将小波对图像结构精细特征的把握能力强的优点与矩所具有的平移,缩放和旋转不变及抗噪性强的特性有机地结合起来,特征向量稳定、识别准确率高、算法快、抗噪性能强,且特征提取方法具有类人视觉特点。
Extracting stabilized and representative feature is the key to character recognition of video image . In this paper , a method of video image character feature extraction is presented based on wavelet and moment analysis . Though seek different moment characters get by the wavelet decomposed sub-graph which got by character image wavelet analysis, to construct character eigenvector. Besides having the invariability to the translation ,scaling and rotation ,this feature vector has the multire solution properties .so it is suitable for classing the very similar objects. Comparing with geometrical moment, the classification rate, recognizing efficiency and antinoise capability of this method have been improved. Furthermore, this feature extraction method has the pattern of human vision.
出处
《微电子学与计算机》
CSCD
北大核心
2003年第11期37-41,共5页
Microelectronics & Computer
关键词
图像字符
字符特征提取
小波变换
矩特征向量
视频图像
模式识别
小波基函数
Character recognition,Pick-up character,Translation scaling and rotation invariant,Wavelet moment eigenvector, Character eigenvector
