摘要
传统的局部二值模式(LBP)算法在特征提取时仅考虑像素区域中心像素点与边缘像素点之间的关系,往往会出现部分重要信息偏失的问题。针对这一不足,提出一种改进的LBP算法,在减少特征信息偏失的同时也提升了特征提取的运算速率。此外,在进行图像分类时对于KNN分类算法存在的训练样本集中近邻样本过多,影响分类结果的问题,也进行了改进。为验证算法的可行性以及优越性,分别用传统LBP算法以及本文改进的LBP算法对图像进行特征提取;并分别使用K最邻近(KNN)算法以及改进KNN算法对图像进行分类,在JAFFE表情库以及CK+表情库上进行实验。结果表明,改进算法在表情识别的准确率以及特征提取的速率上都有很大的提高。
The traditional local binary pattern(LBP)algorithm only considers the relationship between the center pixel and the edge pixel in the pixel region,which often leads to the problem of partial loss of important information.In order to solve this problem,this paper proposes an improved LBP algorithm,which can reduce the bias of feature information and improve the speed of feature extraction.In addition,in the process of image classification,the KNN classification algorithm has the problem that too many neighbor samples in the training sample set affect the classification results.In order to verify the feasibility and superiority of the algorithm in this paper,the traditional LBP algorithm and the improved LBP algorithm are used to extract the image features;KNN algorithm and the improved KNN algorithm are used to classify the images,and the experiments are carried out on the Jaffe expression database and the CK+expression database.The results show that:the improved algorithm has a great improvement in the accuracy of expression recognition and the speed of feature extraction.
作者
贾锋
王高
师钰璋
付雨泽
Jia Feng;Wang Gao;Shi Yuzhang;Fu Yuze(Information and Communication Engineering Institute,North University of China,Taiyuan 030000,China)
出处
《国外电子测量技术》
2020年第8期40-44,共5页
Foreign Electronic Measurement Technology
关键词
表情识别
特征提取
局部二值模式
K近邻
expression recognition
feature extraction
local binary patterns
K nearest neighbor
