CAJ | 학술논문

提出了融合双向2DLDA的Gabor子图像人脸识别方法。将人脸分块子图像进行Gabor特征提取，由于2DLDA提取的主要是人脸图像水平方向上的判别信息，扩展2DLDA提取的主要是人脸图像垂直方向上的判别信息。本文将人脸分块子图像进行Gabor特征提取后映射到水平和垂直两个不同的2DLDA特征子空间，并对这两类特征进行融合，把每一块的Gabor子图像作为局部特征利用支持向量机分类器进行分类与识别，通过将这些局部分类器进行集成，得到最后的决策。在FERET和ORL人脸数据库的实验结果验证了本方法的有效性。
제출료융합쌍향2DLDA적Gabor자도상인검식별방법。장인검분괴자도상진행Gabor특정제취，유우2DLDA제취적주요시인검도상수평방향상적판별신식，확전2DLDA제취적주요시인검도상수직방향상적판별신식。본문장인검분괴자도상진행Gabor특정제취후영사도수평화수직량개불동적2DLDA특정자공간，병대저량류특정진행융합，파매일괴적Gabor자도상작위국부특정이용지지향량궤분류기진행분류여식별，통과장저사국부분류기진행집성，득도최후적결책。재FERET화ORL인검수거고적실험결과험증료본방법적유효성。
Face recognition using sub-images Gabor transform and fusion bidirectional 2DLDA is proposed. we know the feature extracted by 2DLDA is mainly the horizontal discriminate characteristic, whereas the feature extracted by extended 2DLDA is mainly the vertical discriminate characteristic. The face images utilize the convolution of the Gabor filters to extract features, then the images are divided into smaller sub-images, these two 2DLDAs transform these sub-images into different spaces respectively, and two classes of features can be obtained. By fusing the two classes of features, each sub-image is designed as a support vector machine(SVM) classifier, by use of weighed average similarity to make the final decision. Experimental results show that our method can have good recognition performance.