CAJ | 학술논문

光照变化容易使人脸图像的灰度分布不均匀,造成局部对比度差别较大,影响人脸识别的效果。为此本文首先分别对人脸图像进行直方图均衡化和对数变换,接着将处理后的图像进行融合,然后运用PCA算法对人脸图像进行特征提取,最后采用三阶近邻分类法来实现人脸识别。通过对Yale、ORL和FERET人脸库的仿真实验结果表明,该方法在人脸图像光照变化的情况下,能够较好地改善人脸补偿的效果,具有较高的平均识别正确率。
광조변화용역사인검도상적회도분포불균균,조성국부대비도차별교대,영향인검식별적효과。위차본문수선분별대인검도상진행직방도균형화화대수변환,접착장처리후적도상진행융합,연후운용PCA산법대인검도상진행특정제취,최후채용삼계근린분류법래실현인검식별。통과대Yale、ORL화FERET인검고적방진실험결과표명,해방법재인검도상광조변화적정황하,능구교호지개선인검보상적효과,구유교고적평균식별정학솔。
Variable illumination make the grayscale distribution uneven in face images. It will make large difference in local contrast and affect the face recognition results. So this paper takes implementing histogram equalization and logarithmic transformation respectively for the face image at first, then fusing the processed image together, and use PCA algorithm for feature extration in face images, we take 3-neighbor classification to achieve face recognition in the end. Simulation results though the Yale,ORL and FERET face database show that this method in variable illumination circumstances in face image can improve the effect of face compensation and have higher average recognition accuracy.