摘要
为有效解决人脸识别中二维Gabor的维数灾难,线性鉴别分析算法(LDA)的小样本问题和因拍摄不慎造成的图像模糊的问题,提出一种图像去模糊的改进Gabor和最小二乘支持向量机(LSSVM)相结合的新算法。首先用约束最小二乘方(CLS)对模糊的人脸图像去模糊,然后将DLDA和二维Gabor相融合进行降维处理,最后利用训练速度快,泛化能力强的LSSVM进行分类识别。并通过ORL和Yale人脸库来做对比验证,证明了此方法的高效性。
In order to effectively solve the problems of curse dimensionality of 2D Gabor, the small sample size of LDA and image blurring due to accidentally shooting in the face recognition, a new algorithm is proposed based on improved Gabor and LSSVM . The improved Gabor solved the problem of deblurring image. First the blurred facial image is dealed with constrained least squares, then reducing the dimensionality of the feature space is done by fusion of DLDA and 2D Gabor. Because the LSSVM has the advantage of fast training speed and strong generalization ability. Finally, we use the LSSVM for classification and recognition. The efficiency of the proposed method is demonstrated by comparative verification of the Yale and ORL face database.
出处
《电视技术》
北大核心
2015年第24期108-112,共5页
Video Engineering
基金
新疆维吾尔自治区科学基金项目(2015211C257)
