[1]许皓,钱宇华,王克琪,等.高低频通道特征交叉融合的低光人脸检测算法[J].智能系统学报,2024,19(2):472-481.[doi:10.11992/tis.202208034]
　XU Hao,QIAN Yuhua,WANG Keqi,et al.Low-light face detection method based on the cross fusion of high-and low-frequency channel features[J].CAAI Transactions on Intelligent Systems,2024,19(2):472-481.[doi:10.11992/tis.202208034]
点击复制

高低频通道特征交叉融合的低光人脸检测算法

《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷: 19 期数: 2024年第2期 页码: 472-481 栏目: 人工智能院长论坛 出版日期: 2024-03-05
Title:
Low-light face detection method based on the cross fusion of high-and low-frequency channel features
作者:
许皓1,2, 钱宇华1,2,3, 王克琪1, 刘畅1,2, 李俊霞1,2
1. 山西大学 大数据科学与产业研究院, 山西 太原 030006;
2. 山西大学 计算机与信息技术学院, 山西 太原 030006;
3. 山西大学 计算智能与中文信息处理教育部重点实验室, 山西 太原 030006
Author(s):
XU Hao1,2, QIAN Yuhua1,2,3, WANG Keqi1, LIU Chang1,2, LI Junxia1,2
1. Institute of Big Data Science and Industry, Shanxi University, Taiyuan 030006, China;
2. School of Computer and Information Technology, Shanxi University, Taiyuan 030006, China;
3. Key Laboratory of Computational Intelligence and Chinese Informat
关键词:
低光人脸检测; 高低频通道特征; 低光增强; 多尺度特征融合; 计算机视觉; 图像处理; 深度学习; 频率域分析
Keywords:
low-light face detection; features of high-and low-frequency channels; low-light enhancement; multiscale feature fusion; computer vision; image processing; deep learning; frequency domain analysis
分类号:
TP183
DOI:
10.11992/tis.202208034
文献标志码:
2023-12-05
摘要:
低光条件下捕获的人脸图像存在着噪声严重、对比度低等不足，极大影响了现有人脸检测器的准确性，另外，现有的低光图像检测算法欠缺小区域人脸信息的提取能力。为此，提出一种基于深度学习的两阶段人脸检测算法，即利用现有的低光图像增强算法对人脸图像进行增强后再进行检测。为提升检测算法对人脸信息的提取能力，设计一种新型的高低频通道特征交叉融合方法，该方法首先使用高低频通道特征可分离模块分离出不同尺度特征的高低频信息，然后对上述信息进行交叉融合，提升网络提取高频细节信息和低频色域信息的能力，进而提高检测网络的性能。对比试验和消融试验验证了该研究方法的有效性，试验结果表明该方法优于基准方法4.0% mAP。
Abstract:
Face images captured under low-light conditions suffer from significant noise and low contrast, which negatively impact the accuracy of existing face detection systems. In addition, existing low-light image detection algorithms struggle to extract information from small facial areas. To tackle these issues, this paper proposes a two-stage face detection algorithm based on deep learning. This algorithm enhances low-light images before initiating the detection process using an established low-light image enhancement method. The objective is to enhance the ability of the detection method to extract facial information. Thus, a new cross-fusion method of high- and low-frequency channel features is designed. The first step involves using a separable module for high- and low-frequency channel features, enabling the separation of different scale features. These separated features are then cross-fused to improve the ability of the network to extract high-frequency details and low-frequency color information. This, in turn, improves the performance of the detection network. The comparative and ablation experiments validate the effectiveness of the proposed method. The experimental results demonstrate that our method surpasses the baseline method by 4.0% mAP.
参考文献/References:
[1] YUAN Ye , YANG Wenhan , REN Wenqi , et al. UG2+Track 2: A collective benchmark effort for evaluating and advancing image understanding in poor visibility environments[EB/OL]. (2019-04-09)[2022-01-01]. https://doi.org/10.48550/arXiv.1904.04474.
[2] XU Yuanyuan, YAN Wan, YANG Genke, et al. CenterFace: joint face detection and alignment using face as point[J]. Scientific programming, 2020, 2020: 1–8.
[3] YANG Shuo, LUO Ping, LOY C C, et al. WIDER FACE: a face detection benchmark[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 5525-5533.
[4] GUO Chunle, LI Chongyi, GUO Jichang, et al. Zero-reference deep curve estimation for low-light image enhancement[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 1777-1786.
[5] PIZER S M, AMBURN E P, AUSTIN J D, et al. Adaptive histogram equalization and its variations[J]. Computer vision, graphics, and image processing, 1987, 39(3): 355–368.
[6] FARID H. Blind inverse Gamma correction[J]. IEEE transactions on image processing, 2001, 10(10): 1428–1433.
[7] LAND E H, MCCANN J J. Lightness and retinex theory[J]. Journal of the optical society of America, 1971, 61(1): 1–11.
[8] GUO Xiaojie. LIME: a method for low-light image enhancement[C]//Proceedings of the 24th ACM International Conference on Multimedia. New York: ACM, 2016: 87-91.
[9] RAHMAN Z U, JOBSON D J, WOODELL G A. Retinex processing for automatic image enhancement[J]. Journal of electronic imaging, 2004, 13(1): 100–110.
[10] FU Xueyang, ZENG Delu, HUANG Yue, et al. A fusion-based enhancing method for weakly illuminated images[J]. Signal processing, 2016, 129: 82–96.
[11] FU Xueyang, ZENG Delu, HUANG Yue, et al. A weighted variational model for simultaneous reflectance and illumination estimation[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 2782-2790.
[12] WANG Shuhang, ZHENG Jin, HU Haimiao, et al. Naturalness preserved enhancement algorithm for non-uniform illumination images[J]. IEEE transactions on image processing:a publication of the IEEE signal processing society, 2013, 22(9): 3538–3548.
[13] LI Mading, LIU Jiaying, YANG Wenhan, et al. Structure-revealing low-light image enhancement via robust retinex model[J]. IEEE transactions on image processing, 2018, 27(6): 2828–2841.
[14] WEI Chen, WANG Wenjing, YANG Wenhan, et al. Deep retinex decomposition for low-light enhancement[EB/OL]. (2018-08-14)[2022-01-01]. https://arxiv.org/abs/1808.04560.pdf
[15] ZHANG Yonghua, ZHANG Jiawan, GUO Xiaojie. Kindling the darkness: a practical low-light image enhancer[C]//Proceedings of the 27th ACM International Conference on Multimedia. New York: ACM, 2019: 1632-1640.
[16] ZHANG Yonghua, GUO Xiaojie, MA Jiayi, et al. Beyond brightening low-light images[J]. International journal of computer vision, 2021, 129(4): 1013–1037.
[17] 王克琪, 钱宇华, 梁吉业, 等. 局部-全局关系耦合的低照度图像增强[J]. 中国科学:信息科学, 2022, 52(3): 443–460
WANG Keqi, QIAN Yuhua, LIANG Jiye, et al. Local-global coupling relationship based low-light image enhancement[J]. Scientia sinica (informationis), 2022, 52(3): 443–460
[18] GONG Yuqi, YU Xuehui, DING Yao, et al. Effective fusion factor in FPN for tiny object detection[C]//2021 IEEE Winter Conference on Applications of Computer Vision. Waikoloa: IEEE, 2021: 1159-1167.
[19] MEI Y, FAN Y, ZHANG Y, et al. Pyramid attention network for image restoration[J]. International journal of computer vision, 2023, 131(12): 3207–3225.
[20] LIM J S, ASTRID M, YOON H J, et al. Small object detection using context and attention[C]//2021 International Conference on Artificial Intelligence in Information and Communication. Jeju Island: IEEE, 2021: 181-186.
[21] XU A, YAO A, LI A, et al. Auto-FPN: automatic network architecture adaptation for object detection beyond classification[C]//2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2020: 6648-6657.
[22] LUO Yihao, CAO Xiang, ZHANG Juntao, et al. CE-FPN: enhancing channel information for object detection[J]. Multimedia tools and applications, 2022, 81(21): 30685–30704.
[23] ZHANG Tianwen, ZHANG Xiaoling, KE Xiao. Quad-FPN: a novel quad feature pyramid network for SAR ship detection[J]. Remote sensing, 2021, 13(14): 2771.
[24] JIANG Yiqi, TAN Zhiyu, WANG Junyan, et al. GiraffeDet: a heavy-neck paradigm for object detection[EB/OL]. (2022-02-09)[2022-04-04]. https://doi.org/10.48550/arXiv.2202.04256.
[25] CAO Junxu, CHEN Qi, GUO Jun, et al. Attention-guided context feature pyramid network for object detection[EB/OL]. (2020-05-23)[2022-04-04]. https://arxiv.org/abs/2005.11475.
[26] SASAGAWA Y, NAGAHARA H. YOLO in the dark - domain adaptation method for merging multiple models[C]//Proceedings of the European Conference on Computer Vision. [S.l.]: Springer, 2020: 345-359.
[27] WANG Wenjing, YANG Wenhan, LIU Jiaying. HLA-face: joint high-low adaptation for low light face detection[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 16190-16199.
[28] LIANG Jinxiu, WANG Jingwen, QUAN Yuhui, et al. Recurrent exposure generation for low-light face detection[J]. IEEE transactions on multimedia, 2022, 24: 1609–1621.
[29] HUANG Gao, LIU Zhuang, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 2261-2269.
[30] LINDEBERG T. Scale-space theory in computer vision[M]. [S. l. ]: Springer Science & Business Media, 2013.
[31] CHEN Yunpeng, FAN Haoqi, XU Bing, et al. Drop an octave: reducing spatial redundancy in convolutional neural networks with octave convolution[C]//2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2020: 3434-3443.
[32] LIN T Y, MAIRE M, BELONGIE S, et al. Microsoft COCO: common objects in context[C]//European Conference on Computer Vision. Cham: Springer, 2014: 740-755.
[33] DA K. A method for stochastic optimization[EB/OL]. (2014-12-22)[2022-01-01]. https://doi.org/10.48550/arXiv.1412.6980.
[34] GIRSHICK R. Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision. Santiago: IEEE, 2016: 1440-1448.
[35] NAJIBI M, SAMANGOUEI P, CHELLAPPA R, et al. SSH: single stage headless face detector[C]//2017 IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 4885-4894.
[36] DENG Jiankang, GUO Jia, VERVERAS E, et al. RetinaFace: single-shot multi-level face localisation in the wild[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 5202-5211.
[37] TANG Xu, DU D K, HE Zeqiang, et al. PyramidBox: A context-assisted single shot face detector[C]//European Conference on Computer Vision. Cham: Springer, 2018: 812-828.
[38] LI Jian, WANG Yabiao, WANG Changan, et al. DSFD: dual shot face detector[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2020: 5055-5064.
[39] LIN T Y, DOLL?R P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 936-944.
[40] CHEN Qiang, WANG Yingming, YANG Tong, et al. You only look one-level feature[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 13034-13043.

备注/Memo

收稿日期:2022-08-22。
基金项目:科技创新2030?“新一代人工智能”重大项目(2021ZD0112400)；国家自然科学基金项目(62136005)；山西省揭榜挂帅项目(202201020101006)
作者简介:许皓，硕士研究生，主要研究方向为夜间人脸检测、目标检测、低光增强。E-mail：552420610@qq.com;钱宇华，教授、博士生导师，山西大学大数据科学与产业研究院院长，主要研究方向为人工智能、大数据、AI for Science。先后主持国家自然科学基金重点项目、国家重点研发计划项目、国防重点项目等国家/省部级项目20余项。曾获山西省科学技术奖（自然科学类）一等奖，入选全球高被引科学家。发表学术论文200余篇。E-mail：jinchengqyh@126.com;王克琪，博士研究生，主要研究方向为深度学习、计算机视觉、低照度图像增强。E-mail：wkq_2021@163.com
通讯作者:钱宇华. E-mail: jinchengqyh@126.com

更新日期/Last Update: 1900-01-01
Copyright © 《 智能系统学报》 编辑部
地址：（150001）黑龙江省哈尔滨市南岗区南通大街145-1号楼 电话：0451- 82534001、82518134　邮箱：tis@vip.sina.com