一种改进的局部三值模式的人脸识别方法

Face recognition based on improved local ternary patterns

为了更好的描述人脸特征,提出了一种基于不同尺度像素块及自适应阈值的局部三值(LTP)模式方法.该方法首先将图像分为若干个子区域,采用自适应阈值并基于不同尺度的像素块提取每个子区域的LTP纹理直方图,然后将得到的每个子区域的直方图连在一起并经过主成分分析(PCA)降维处理得到特征向量.在人脸数据库上进行的实验证明,应用该方法进行人脸特征提取并结合最近邻分类法得到了较高的识别率.

In order to describe the facial feature,a local ternary patterns（LTP）method based on different scale pixel blocks and adaptive thresholds was proposed.Firstly,the image was divided into several sub regions.The LTP texture histogram of each sub region was extracted by using the adaptive threshold based on pixel blocks with different scales.Then the histogram of each sub region was connected to obtain the final feature vector by the principal component analysis（PCA）.Through experiments on the face database,a higher recognition rate was obtained.