摘要： 使用小波对人脸图像进行不同尺度的分解，获得对应的局部相位量化特征，结合模糊理论并融合各尺度下测试样本的差异隶属度，提出一种模糊多尺度局部相位量化的人脸识别方法。在ORL和FERET数据库上进行实验，结果表明，该方法的识别率较高，在提取更多人脸特征的同时，能避免传统多尺度方法中容易造成的高维问题，对光照和噪声具有更高的鲁棒性

关键词: 人脸识别, 多尺度分析, 局部相位量化, 模糊理论, 直方图, 特征抽取

Abstract: The face images are decomposed using multi-level wavelet transformation to get the Local Phase Quantization(LPQ) features. The membership grades of the test images to the training images under different level are fused based on the fuzzy theory. This paper proposes a face recognition method by fuzzy multi-scale LPQ. Doing experiments on the ORL and FERET face database, results show that this method can get high recognition rate, not only effectively extracts the face feature but also solves the problem of high-dimensional feature that the original multi-level method brings. It shows better robustness to non-uniform illumination and noise.

Key words: face recognition, multi-scale analysi, Local Phase Quantization(LPQ), fuzzy theory, histogram, feature extraction

中图分类号: 

• TP18