摘要
在EBGM算法中,计算人脸特征点的相似度时,认为每个Gabor小波系数对结果的影响是相同的,从而给予所有系数同等权重,但实验测试表明情况并非如此。充分挖掘了人脸的统计特性,通过对不同特征点在频域分布特征的提取,提出一种对Gabor小波系数进行分类并赋予其不同权重的优化方法。实验测试证明,该算法能有效地提高识别率。
In elastic bunch graph matching（EBGM）algorithm, all the Gabor wavelet parameters have the same weight when measuring the similarity between two jets of the same landmark. In this paper, by making use of the statistics of human face and extracting frequency distribution of different landmarks, a new similarity measuring method is presented, based on the optimized weight of Gabor wavelet parameters. The experiments show that this method can effectively improve the recognition rate.
出处
《信息技术》
2009年第1期59-62,共4页
Information Technology
