摘要
充分考虑到人脸对称性,结合对称主量分析和Fisher线性判别,提出了改进的Fisherface方法——对称Fisherface(SFisherface).在人脸的表情、姿态以及光照发生变化的情况下,通过实验说明了SFisherface在性能上面的优势.通过比较镜像扩展样本后的Fisherface、SFisherface和经典Fisherface方法的联系和区别,说明了SFisherface有助于提取稳定性的特征,并能够显著提高识别率.
By combining symmetrical PCA with Fisher linear discrimination, symmetrical Fisherface （SFisherface） was proposed. Experiments were conclucted to prove the effectiveness of SFisherface by varying the illumination, pose and expression. The method using mirror images for sample expansion was compared with SFisherface and Fisherface, and the conclusions show that SFisherface can find more stable features and help raise the recognition rate remarkably.
