CAJ | 학술논문

传统的LBP算子只利用了局部的信息，而忽略了全局信息。MB_LBP算子虽然充分考虑了全局信息，但对局部信息的表示不足。在此提出一种改进后的LBP特征的人脸识别方法，改进后的LBP算子不仅能够利用局部特征，同时也兼顾了全局信息。该方法首先将人脸图像分块，对于每个分块，计算LBP特征，对于得到的LBP特征，根据其中心像素和分块灰度均值关系重新进行计算得到改进后的LBP特征，最后采用最近邻分类器进行识别。在ORL和YALE标准人脸数据库上的实验表明，改进后的识别效果优于使用传统LBP算子和MB_LBP算子。采用改进后的LBP算子，能够明显提高识别率，在ORL和YALE的实验显示能提高3%~8%左右的识别率。
전통적LBP산자지이용료국부적신식，이홀략료전국신식。MB_LBP산자수연충분고필료전국신식，단대국부신식적표시불족。재차제출일충개진후적LBP특정적인검식별방법，개진후적LBP산자불부능구이용국부특정，동시야겸고료전국신식。해방법수선장인검도상분괴，대우매개분괴，계산LBP특정，대우득도적LBP특정，근거기중심상소화분괴회도균치관계중신진행계산득도개진후적LBP특정，최후채용최근린분류기진행식별。재ORL화YALE표준인검수거고상적실험표명，개진후적식별효과우우사용전통LBP산자화MB_LBP산자。채용개진후적LBP산자，능구명현제고식별솔，재ORL화YALE적실험현시능제고3%~8%좌우적식별솔。
The traditional LBP(local binary pattern)operator can capture the local information only,but ignores global in?formation. MB_LBP operator expresses local information insufficiently,though it fully considers global information. A new face recognition algorithm based on an improved LBP operator is proposed in this paper. The improved LBP operator can capture not only the local features,but also the global features. In this method,the original face image is divided into small blocks,and then the LBP features of each block are calculated. According to the relation of center pixel and gray mean value of each block, the obtained LBP features are re?calculated to get the improved LBP features. Finally,the classification is performed by a nearest neighbor classifier. The result of the experiment with ORL and YALU standard face database shows that the recognition efficien?cy of the improved LBP operator is better than traditional LBP operator and MB?LBP operator,the recognition rate is increased by 3%~8%. The application of the improved LBP operator can improve the recognition efficiency obviously.