空间加权伪Zernike矩的人脸图像特征提取方法

Face image feature extraction method based on weighted pseudo-Zernike moments

为了解决人脸图像检索中人脸定位不精确的问题,提出了一种空间加权伪Zernike矩的人脸图像特征提取方法。该方法通过设计的加权函数来强调或突出面部上的关键部位,如眼睛、鼻子、嘴巴等区域的重要性,因而能够对人脸特征进行更精确地提取。仿真实验结果表明,利用空间加权伪Zernike矩进行人脸区域描述时,检索准确率明显优于传统的主成分分析方法和直接伪Zernike矩特征提取方法。

Human face image retrieval face the problem of inaccurate positioning,a space-weighted pseudo-Zernike moments face image feature extraction methods is presented. This method uses the weighting function to emphasize or highlight the key parts of the face such as eyes,nose,mouth,such as the importance of the region and therefore able to carry out facial feature extraction more accurate. The simulation results show that the use of space-weighted pseudo-Zernike moments to describe the facial region,the retrieval accuracy is better than the traditional method of principal component analysis and direct pseudo-Zernike moment feature extraction methods.