中文摘要：

为了解决局部二值模式识别率不高、特征向量维数大的问题,使用Kirsch边缘检测算子,根据像素水平、垂直及对角的3个方向响应进行编码,得到基于Kirsch算子的局部二值模式.该方法有效提取了面部表情特征,且维数低,适合在大型人脸库上运行.在JAFFE数据库上的实验结果表明,基于Kirsch算子的局部二值模式与局部二值模式、CBP、GDP及LDP相比,识别率和运行效率均有提高,是一种有效的特征描述算子.

英文摘要：

In order to address the problem of local binary pattern of low discrimination and high dimension, this paper presents a Kitsch local binary pattern operator which tests operators by using Kirsch operator and encodes according to horizontal, vertical and diagonal responses of pixels. This method is applicable to large-scale face database because it effectively extracts the facial features with a low dimension. The tests in JAFFE database show that compared with local binary pattern, CBP, GDP and LDP, Kirsch-operator-based local binary pattern is improved in recognition rate and efficiency, and is thus an effective operator for characterization.