基于级联SVM的无参考人脸图像质量评价系统  

Non-reference face image quality evaluation system based on cascaded SVM

作　　者：李昆仑[1] 熊婷[1] 张炘[1] 廖频 LI Kunlun;XIONG Ting;ZHANG Xin;LIAO Pin(School of Science and Technology,Nanchang University,Nanchang 330029,China;School of Information Engineering,Nanchang University,Nanchang 330029,China)

机构地区：[1]南昌大学科学技术学院,江西南昌330029 [2]南昌大学信息工程学院,江西南昌330029

出　　处：《现代电子技术》2018年第24期98-100,105,共4页Modern Electronics Technique

基　　金：江西省教育厅科学技术研究项目(GJJ161525;GJJ151510;GJJ151506);江西省科技厅研究项目(20161BBE50049)~~

摘　　要：针对低质量人脸图像严重影响人脸检测与分类精度的问题,提出一种基于级联SVM的无参考人脸图像质量评价系统。该系统从人脸图像的分类性能出发,使用不同程度的模糊、光照不均和含噪声退化的人脸图像数据集训练级联SVM,并将SVM的输出作为图像质量的评分。仿真测试与实验结果表明,所提出的系统能较好处理各种退化情况,相比于传统的图像质量评估方式能获得更准确的评估分数。In allusion to the problem that the low-quality face image seriously affects face detection and classification accu-racy,a non-reference face image quality evaluation system based on cascaded SVM is proposed.The datasets of face images with different degrees of blur,uneven illumination and noise degradation are used to train the cascaded SVM according to the classifi-cation performance of face images.The output of the SVM is taken as the evaluation score of the image quality.The simulation test and experimental results show that the proposed system can deal with various degradation situations well,and obtain more accurate evaluation scores in comparison with the traditional image quality evaluation method.

关 键 词：级联SVM 人脸检测 图像识别 质量评价系统 分类性能 噪声退化 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TM76[电子电信—信息与通信工程]