基于衍生小波的笔迹鉴别特征的表示与提取

项目名称： 基于衍生小波的笔迹鉴别特征的表示与提取

项目编号： No.60803056

项目类型： 青年科学基金项目

立项/批准年度： 2009

项目学科： 轻工业、手工业

项目作者： 何震宇

作者单位： 华中科技大学

项目金额： 20万元

中文摘要： 笔迹鉴别是通过分析、比较笔迹图像的特征来研究人体行为特征并进一步进行书写者的身份鉴别与验证的的一门学科，能广泛地应用到需要安全控制的各个领域，具有重要的社会、经济价值。笔迹特征的提取与表示是笔迹鉴别理论研究的核心之一，在本项研究中我们将集中研究如何利用衍生小波的方向分析、线奇异性检测等优势，根据手写笔迹图像的具体特点， 选择、改进已有的衍生小波来表示与提取笔迹特征，构造新的衍生小波与相关滤波器来分析与提取笔迹特征，侧重于表示与提取笔迹的纹理特征以及笔迹轮廓曲率特征，解决笔迹特征表示与提取中长期未能得到解决的一些重要理论问题，并进一步建立表示与提取这两种特征的相应实现算法，为模式识别与图像处理中特征表示与提取提供一些新的理论与方法。

中文关键词： 笔迹鉴别；小波分析；衍生小波；笔迹纹理特征；笔迹轮廓曲率特征

英文摘要： Writer identification is a discipline to identify or verify the writer’ status by analyzing and matching the features of handwritten images. It can be widely used in various fields where security control is needed, and is of high economical and social values. Writer identification generally consists of several steps: pre-processing, feature extraction, similarity measurement and performance evaluation. Feature extraction is the core one among these steps. In this research, with respect to the properties of handwritten images and merits of wavelet-like frameworks on directional analysis and detection of linear singularity, we will concentrate on how to choose and optimize the existing wavelet-like frameworks or construct new wavelet-like frameworks and corresponding filters to characterize and extract the features of handwritten images, especially the texture feature and the contour curvature feature. Our research can develop new algorithms to characterize and extract the two above-mentioned features of handwritten images, solve some important unsolved theoretical problems of write identification as well as provide some new theories and methodologies for the fields of image process and pattern recognition.

英文关键词： write identification; wavelet analysis; wavelet-like frawework; texture feature; contour curvature feature