自然场景中小样本人脸图像修复对人脸识别的影响

doi:10.3969/j.issn.1671-2072.2024.05.011

中国司法鉴定 ›› 2024 ›› Issue (5): 80-85.DOI: 10.3969/j.issn.1671-2072.2024.05.011

自然场景中小样本人脸图像修复对人脸识别的影响

李胜男^1，2，杨梦暄^1，2，曾锦华^1，2

1.成都信息工程大学软件工程学院，四川成都 610225；
2.司法鉴定科学研究院上海市司法鉴定专业技术服务平台司法部司法鉴定重点实验室，上海 200063

收稿日期:2023-09-11 出版日期:2024-09-15 发布日期:2024-11-28
通讯作者: 曾锦华（1985—），男，高级工程师，博士，主要从事模式识别、多媒体取证以及声像资料和电子数据司法鉴定技术研究工作。E-mail：zengjh@ssfjd.cn
作者简介:李胜男（1999—），女，硕士研究生，主要从事声像资料研究。E-mail：aspirant2022@yeah.net
基金资助:
上海市科委技术标准项目（21DZ2200100）；中央级科研院所公益研究基金（GY2021G-3）；上海市司法鉴定专业技术服务平台资助项目

The Influence of Image Restoration of Small Face Sample in Natural Scenes on Face Recognition

LI Shengnan^1,2, YANG Mengxuan^1,2, ZENG Jinhua^1,2

1. School of Software Engineering, Chengdu University of Information Technology, Chengdu, 610225, China;
2. Shanghai Forensic Service Platform, Key Laboratory of Forensic Science, Ministry of Justice, Academy of Forensic Science, Shanghai 200063, China

Received:2023-09-11 Published:2024-09-15 Online:2024-11-28

摘要/Abstract

摘要： 目的旨在研究小样本人脸图像的质量及修复方法对人脸识别系统性能的影响，以提升司法鉴定识别效果。方法研究自然场景中小样本人脸图像修复对人脸识别技术的影响，小样本人脸图像从Labeled Faces in the Wild数据库中选取，并采用高斯滤波技术进行处理以模拟真实人脸图像质量。同时，使用基于深度神经网络的不同修复方法对模糊人脸图像进行高清修复，通过统计量化分析方法研究不同小样本人脸图像质量和修复方法对人脸识别系统识别性能的影响。结果人脸图像修复技术可以极大地提高小样本人脸图像的清晰度和身份信息视觉辨识度，然而，修复后的人脸图像在人脸识别系统中却产生了负面作用。结论在人像鉴定中，人脸图像修复技术具有积极作用，但在人脸识别技术应用中，应谨慎使用。

关键词: 人脸图像修复, 人脸识别, 人像鉴定, 小样本人脸

Abstract: Objective The aim is to investigate the impact of the quality and image restoration methods of small face sample on the performance of facial recognition systems, in order to improve the effectiveness of forensic appraisal recognition. Methods This paper focused on the impact of few-shot face image inpainting on face recognition technology in natural scenes. We selected few-shot face images from the Labeled Faces in the Wild database. Then, we used Gaussian filtering technology on these images. This simulated the quality of real-life blurred face images. At the same time, different face image inpainting methods based on deep neural networks were used to repair the blurred face images. The influence of different few-shot face images quality and different inpainting methods on the recognition performance of face recognition was studied by statistical analysis. Results The face image inpainting technology enhanced the clarity of small face image samples. Moreover, the technology boosted the visual recognition of identity information in these images. However, the repaired face images negatively affected face recognition. Conclusion Face image inpainting technology plays a positive role in portrait identification, but it should be used more carefully when be applied in face recognition.

Key words: face image inpainting, face recognition, portrait identification, small face sample

中图分类号:

TP309

李胜男, 杨梦暄, 曾锦华. 自然场景中小样本人脸图像修复对人脸识别的影响[J]. 中国司法鉴定, 2024(5): 80-85.

LI Shengnan, YANG Mengxuan, ZENG Jinhua. The Influence of Image Restoration of Small Face Sample in Natural Scenes on Face Recognition[J]. Chinese Journal of Forensic Sciences, 2024(5): 80-85.

参考文献

[ 1 ] JAIN A，HONG L，PANKANTI S. Biometric identification[J]. Communications of the ACM，2000，43（2）：90-98.
[ 2 ] KAUR N. A study of biometric identification and verification system[C]//Proceedings of 2021 International Conference on Advance Computing and Innovative Technologies in Engineering （ICACITE）. Washington D.C.，USA：IEEE Press，2021：60-64.
[ 3 ] LI L X，MU X H，LI S Y，et al. A review of face recognition technology[J]. IEEE Acess，2020，8：139110-139120.
[ 4 ] CHEN L F，LIAO H M，KO M，et al. A new LDA-based face recognition system which can solve the small sample size problem[J]. Pattern Recognition，2000，33（10）：1713-1726.
[ 5 ] LI Z，LIU F，YANG W J，et al. A survey of convolutional neural networks：analysis，applications，and prospects[J]. IEEE Transactions on Neural Networks and Learning Systems，2021，33（12）：6999-7019.
[ 6 ] GU J X，WANG Z H，KUEN J，et al. Recent advances in convolutional neural networks[J]. Pattern Recognition，2018，77：354-377.
[ 7 ] KIM P. Convolutional neural network[J]. MATLAB Deep Learning，2017，2：121-147.
[ 8 ] GOODFELLOW I，POUGET-ABADIE J，MIRZA M，et al. Generative adversarial networks[J]. Communications of the ACM，2020，63（11）：139-144.
[ 9 ] MIRZA M，OSINDERO S. Conditional generative adversarial nets[J]. Computer Science：1411.1784，2014.
[10] SAHARIA C，HO J，CHAN W，et al. Image super-resolution via iterative refinement[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence，2022，45（4）：4713-4726.
[11] KNOCHE M，HORMANN S，RIGOLL G. Cross-quality LFW：A database for analyzing cross-resolution image face recognition in unconstrained environments[C]//Proceedings of 2021 16th IEEE International Conference on Automatic Face and Gesture Recognition （FG 2021）.Washington D.C.，USA：IEEE Press，2021：1-5.
[12] BHARATI S，KHAN T Z，PODDER P，et al. A comparative analysis of image denoising problem：noise models，denoising filters and applications[J]. Cognitive Internet of Medical Things for Smart Healthcare：Services and Applications，2021：49-66.
[13] QIN Z，ZENG Q，ZONG Y，et al. Image inpainting based on deep learning：A review[J]. Displays，2021，69：102028.
[14] JOSHUA H.HitPaw Photo Enhancer—the best AI photo quality enhancer[EB/OL].[2023-09-01]. https：//www.hitpaw.com/photo-enhancer.html.
[15] Topaz Labs. Topaz Photo AI—maximize image quality with AI[EB/OL]. [2023-09-19].https：//www.topazlabs.com/topaz-photo-ai.
[16] Aliyun. Human face—Alibaba cloud visual intelligence open platform[EB/OL][2023-05-01].https：//vision.aliyun.com/facebody.
[17] Baidu.Recognition cloud service[EB/OL].[2023-06-01].https：//cloud.baidu.com/product/face.
[18] ArcSoft Technology Co.，Ltd.ArcSoft vision open platform[EB/OL].[2023-05-01].https：//ai.arcsoft.com.cn/product/arcface.html.

[1]	卢启萌, 毛晓, 凌嵘, 陈晓红, 曾锦华, 耿浦洋, 孙维龙, 卞新伟, 施少培. 口罩佩戴对人像鉴定的影响 [J]. 中国司法鉴定, 2021, 0(5): 89-94.
[2]	曾锦华施少培卞新伟卢启萌邱秀莲. 人脸识别技术在人像鉴定中的应用研究[J]. 中国司法鉴定, 2019, 0(2): 89-93.
[3]	曾锦华, 施少培, 奚建华, 李岩, 卢启萌. 人像鉴定中的头皮头发特征应用[J]. , 2016, 0(5): 32-36.
[4]	卢启萌, 施少培, 李岩, 曾锦华. 一种基于动态比对表的步态特征检验方法[J]. , 2015, 0(3): 61-64.

自然场景中小样本人脸图像修复对人脸识别的影响

The Influence of Image Restoration of Small Face Sample in Natural Scenes on Face Recognition

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 4

编辑推荐

Metrics