基于优化Gabor滤波器和GMRF的笔迹特征提取方法

Handwriting feature extraction method based on optimal Gabor filter and GMRF

提取有效的特征一直是笔迹鉴别的关键问题,针对传统Gabor滤波器特征提取方法存在的不足,充分利用Gabor滤波系数间的相关关系,提出一种融合全局特征和局部特征的特征提取方法。该方法先通过字符笔画的方向梯度直方图(HOG)来优化Gabor滤波器的角度参数,再利用高斯马尔科夫随机场(GMRF)模型对Gabor滤波图像中的不同局部结构信息进行描述,最终得到笔迹图像的整体特征。以楷书四大家的真迹样本和收集的英文手稿作为实验数据,采用最小加权欧式距离分类器对笔迹样本进行分类,通过五重交叉验证法分别得到97.6%和88.3%的正确分类率,表明该方法提取的特征具有较强的笔迹表征能力,是一种有效的笔迹特征提取方法。

Extracting effective features to describe handwriting is always a key problem in writer identification. In order to overcome the shortcomings of the traditional Gabor filter method, as well as to fully exploit correlation between Gabor filtering coefficient, this paper proposes a novel method for handwriting feature extraction, which merges the global and local features together. Histogram Of Gradient（HOG）of the character strokes is firstly used to optimize the orientations of Gabor filter, then Gauss Markov Random Field（GMRF）models are developed for every filtered image to describe the different local spatial structures, and finally it obtains the overall style characteristics of the handwriting images. With the four most famous regular script writers＇ original samples and the collected English scripts as the experimental data, the minimum weighted Euclidean distance classifier is applied to classify handwriting samples, respectively achieving correct classification rates of 97.6% and 88.3% with five-fold cross validation method, which shows that the extracted features have strong ability to characterize the handwriting, and the proposed method is effective in handwriting feature extraction.