【摘要】 光照变化是影响现有人脸识别算法性能的主要因素之一.基于边缘特征的方法能获得较好的光照鲁棒性,且易于实现,但它对表情变化的鲁棒性较差.本文提出了一种融合二值边缘特征和灰度特征的人脸识别方法,并首次将二阶互信息相似性测度引入人脸识别中.在AR图像集和Yale图像集上的实验表明,本方法对含有光照变化和表情变化的图像能获得比现有其它方法更好的总体识别率,具有较好的实用价值.更多还原

【Abstract】 Illumination change is of great importance in affecting the performance of some existing face recognition algorithms.Though edge-based methods are robust to illumination variation and are easy to implement,they do not work perfectly in the cases with expression variation.In order to improve both the lighting robustness and expression robustness,a novel face recognition method based on the fusion of binary edge and grayscale features was proposed.Also the second-order mutual information was used for the similarity metric of grayscale face image for the first time.AR dataset and Yale dataset with various illumination and expression variation were tested to evaluate the effect of the proposed method.Results showed that the overall face recognition rate of the proposed method was better than that of other methods.And these results indicate that our method is more effective for practical use.更多还原