CAJ | 학술논문

提出了一种基于二维多尺度局部二进制模式的人脸识别方法，对一幅人脸图像进行分块，对每一块的图像进行MB-LBP（Multi-scale Block Local Binary Patterns）算子运算，将MB-LBP与灰度共现矩阵结合起来得到了可以更好地描述局部纹理空间结构的二维MB-LBP特征，将各子块的二维MB-LBP特征进行连接形成人脸特征。该算法在ORL和CMU-PIE人脸数据库上进行测试，选择了支持向量机（SVM）作为分类器，并与传统的基于一维LBP特征进行比较，结果表明提出的算法在人脸识别问题上的有效性和优越性。
제출료일충기우이유다척도국부이진제모식적인검식별방법，대일폭인검도상진행분괴，대매일괴적도상진행MB-LBP（Multi-scale Block Local Binary Patterns）산자운산，장MB-LBP여회도공현구진결합기래득도료가이경호지묘술국부문리공간결구적이유MB-LBP특정，장각자괴적이유MB-LBP특정진행련접형성인검특정。해산법재ORL화CMU-PIE인검수거고상진행측시，선택료지지향량궤（SVM）작위분류기，병여전통적기우일유LBP특정진행비교，결과표명제출적산법재인검식별문제상적유효성화우월성。
This paper proposes a novel face recognition method based on two-dimensional multi-block local binary pat-tern. A face image is divided into some blocks. The MB-LBP operator is applied on each block. It combines the MB-LBP and gray level co-occurrence matrix to get two-dimensional MB-LBP features which better describe the local texture spatial structure. Each sub-block of LBP gray histogram is connected to form the face feature. The proposed approach is tested on ORL and CMU-PIE face database with the Support Vector Machine(SVM)as classifier. The results of experiments dem-onstrate that the proposed algorithm is more steady and superior of identifying faces while compared with the traditional one-dimensional LBP features.