中文摘要：

针对局部方向模式（LDP）采样不充分和对噪声敏感的缺点，提出了一种改进的LDP（ILDP）人脸描述方法，利用局部井型邻域的梯度信息描述人脸特征。首先，将中心像素的井型邻域根据采样半径分成两个3×3的子邻域，每个子邻域包含按照原来相对位置排列的8个像素；然后，将两个子邻域与Kirsch模板卷积分别得到两组边缘梯度值，ILDP仅使用两组梯度值中各自最大值的方向编码成一个二位八进制数，产生ILDP码；最后，在人脸描述阶段将人脸图像进行分块并把每块转换成ILDP图，再对ILDP图进行直方图统计，将所有子块的直方图连接生成人脸特征。实验结果表明，ILDP比其它同类基于局部纹理特征的单一人脸描述器在对抗随机噪声方面更具鲁棒性。

英文摘要：

For obtaining more extensive sampling information, this paper presents a novel approach based on improved local directional pattern （ILDP） for face recognition,which adopts the edge gradient information of local #-shape neighborhood including 16 pixels. Specially,the local g/-shape neighborhood of each pixel is equally divided into two 3 X 3 sub-neighborhoods, and 8 edge gradient values of each subneighborhood are gained by convolving it with 8 Kirsch masks respectively. The ILDP just utilizes the di- rection of the largest edge gradient value of each sub-neighborhoo＆ Then, these two directions are enco- ded into a double-digit octal number to produce the ILDP code. Finally,the face descriptor is represented by using the global concatenated histogram based on ILDP map extracted from the face image which is divided into several sub-btocks. The experimental results demonstrate that the proposed method in this paper is more robust to random noise than other single face descriptors based on local texture informa- tion.