基于非下采样轮廓波变换的离线签名识别

Off-line signature recognition based on non-downsampled contourlet transform

为提高手写签名的识别率,提出一种基于NSCT子带纹理特征融合的签名识别方法。对签名图像进行预处理(包括灰度化、平滑、二值化、归一化、细化等),对签名图像进行非下采样Contourlet变换,对变换产生的子带分别提取多级区域局部二值模式和灰度共生矩阵特征,通过融合形成新特征。数据库包含维吾尔文和柯尔克孜文两类文种,每个文种100人(20个样本/人),共4000个签名样本进行实验,实验结果表明,该方法能更准确地提取签名图像多尺度、多方向的纹理特征,可有效提高识别率。

To improve the recognition rate of handwritten signature,a signature recognition method based on NSCT sub-band texture feature fusion was proposed.The signature image was preprocessed(including grayscale,smoothing,binarization,normalization,refinement,etc.).The non-subsampled Contourlet transform was performed on the signature image,and the multi-level regional local binary pattern and the gray level co-occurrence matrix feature were extracted respectively for the sub-bands generated by transformation,and new features were formed by fusion.The database contained Uyghur and Kirgiz languages,each of which had 100 people(20 samples/person)and a total of 4000 signature samples.Experimental results show that the proposed method can extract signature images more accurately.Multi-directional texture features can effectively improve the re-cognition rate.