面部图像纹理特征检测方法研究

The Research on Texture Feature Detection Method of Facial Image

传统面部图像的特征检测主要针对于灰度图像和二值图像,目前国内对于面部彩色图像(尤其是高清面部彩色图像)纹理特征的研究还处于起步阶段。针对于面部区域彩色图像特征,通过采集不同测试者早(健康状态)、晚(疲劳状态)大量面部彩色图像,据此建立面部图像数据库,划分数据库为P(疲劳图)、Z(正常图)、R(结果)三类。利用共生矩阵算法求得最大特征区域的纹理特征向量,用对比统计得到的数据测定正常与疲劳面部图像特征值的差别。实验结果表明,面部彩色图像的纹理特征反映了图像本身的属性,进一步描述了图像的细节信息,具有计算量小的特点。

Traditional facial image feature detection is mainly aimed at gray image and binary image, the current domestic for fa- cial color images, particularly hd facial color image texture characteristics of the research is still in a lower level. This research is mainly aimed at to the face region color ~mage characteristics. By collecting different tester early （health） and late （fatigue） a large number of facial color image, on the basis of establishing the face image database, database is divided into P （fatigue）, Z （normal）, R （results） three categories. Based on the symbiotic matrix algorithm for the biggest regional characteristics of the texture feature vector, use contrast statistical data to determine the difference of normal and fatigue facial image char- acteristic value. Experimental results show that the facial color image texture characteristics reflect the properties of the image itself, further describes the image details, has the characteristics of small amount of calculation.