Journal of Computer Applications ›› 2021, Vol. 41 ›› Issue (11): 3178-3184.DOI: 10.11772/j.issn.1001-9081.2020121894

• Artificial intelligence • Previous Articles     Next Articles

Multiply distortion type judgement method based on multi-scale and multi-classifier convolutional neural network

Junhua YAN1,2(), Ping HOU1,2, Yin ZHANG1,2, Xiangyang LYU1,2, Yue MA1,2, Gaofei WANG1,2   

  1. 1.Key Laboratory of Space Photoelectric Detection and Perception of Ministry of Industry and Information Technology (Nanjing University of Aeronautics and Astronautics),Nanjing Jiangsu 211106,China
    2.College of Astronautics,Nanjing University of Aeronautics and Astronautics,Nanjing Jiangsu 211106,China
  • Received:2020-12-04 Revised:2021-07-25 Accepted:2021-08-03 Online:2021-07-25 Published:2021-11-10
  • Contact: Junhua YAN
  • About author:YAN Junhua,born in 1972,Ph. D.,professor. Her research interests include image quality assessment, multi-source information fusion,target detection,tracking and recognition
    HOU Ping,born in 1993,M. S. candidate. His research interests include image quality assessment
    ZHANG Yin, born in 1986, Ph. D., lecturer. His research interests include simulation and processing of photoelectric detection information
    LYU Xiangyang,born in 1996,M. S. candidate. His research interests include image quality assessment
    MA Yue,born in 1991,M. S. candidate. His research interests include image quality assessment
    WANG Gaofei,born in 1996,M. S. candidate. Her research interests include super-resolution reconstruction of remote sensing images.
  • Supported by:
    the National Natural Science Foundation of China(61705104);the Fundamental Research Funds for the Central Universities(NJ2020021);the Natural Science Foundation of Jiangsu Province(BK20170804)

基于多尺度多分类器卷积神经网络的混合失真类型判定方法

闫钧华1,2(), 侯平1,2, 张寅1,2, 吕向阳1,2, 马越1,2, 王高飞1,2   

  1. 1.空间光电探测与感知工业和信息化部重点实验室(南京航空航天大学),南京 211106
    2.南京航空航天大学 航天学院,南京 211106
  • 通讯作者: 闫钧华
  • 作者简介:闫钧华(1972—),女,陕西兴平人,教授,博士,主要研究方向:图像质量评价、多源信息融合、目标检测、跟踪与识别
    侯平(1993—), 男,江苏靖江人,硕士研究生,主要研究方向:图像质量评价
    张寅(1986—),男,江苏镇江人,讲师,博士,主要研究方向:光电探测信息仿真与 处理
    吕向阳(1996—),男,河北景县人,硕士研究生,主要研究方向:图像质量评价
    马越(1991—),男,江苏淮安人,硕士研究生,主要研究 方向:图像质量评价
    王高飞(1996—),女,河南新野人,硕士研究生,主要研究方向:遥感图像超分辨率重建。
  • 基金资助:
    国家自然科学基金资助项目(61705104);中央高校基本科研业务费专项基金资助项目(NJ2020021);江苏省自然科学基金资助项目(BK20170804)

Abstract:

It is difficult to judge the image multiply distortion type. In order to solve the problem, based on the idea of deep learning multi-label classification, a new multiply distortion type judgement method based on multi-scale and multi-classifier Convolutional Neural Network (CNN) was proposed. Firstly, the image block containing high-frequency information was obtained from the image, and the image block was input into the convolution layers of different receptive fields to extract the shallow feature maps of the image. Then, the shallow feature maps were input into the structure of each sub-classifier for deep feature extraction and fusion, and the fused features were judged by the Sigmoid classifier. Finally, the judgment results of different sub-classifiers were fused to obtain the multiply distortion type of image. Experimental results show that, on the Natural Scene Mixed Disordered Images Database (NSMDID), the average judgment accuracy of the proposed method can reach 91.4% for different types of multiply distortion types in the images, and most of them are above 96.8%, illustrating that the proposed method can effectively judge the types of distortion in multiply distortion images.

Key words: multiply distortion type, multi-label classification, Convolutional Neural Network (CNN), high-frequency information, Sigmoid classifier

摘要:

针对图像混合失真类型判定难的问题,在深度学习多标签分类思想的基础上,提出了一种基于多尺度多分类器卷积神经网络(CNN)的混合失真类型判定方法。首先,从图像中截取得到含有高频信息的图像块,将该图像块输入到不同感受野的卷积层中以提取图像的浅层特征图;其次,将浅层特征图输入到各子分类器结构中以进行深层次的特征提取和融合,将融合的特征通过Sigmoid分类器得到判定结果;最后,将各子分类器的判定结果进行融合得到图像的混合失真类型。实验结果表明,在自然场景混合失真数据库(NSMDID)上,所提方法对图像中存在的混合失真类型的平均判定准确率可以达到91.4%,且对大部分类型的判定准确率都在96.8%以上,可见所提方法能够对混合失真图像中的失真类型进行有效的判定。

关键词: 混合失真类型, 多标签分类, 卷积神经网络, 高频信息, Sigmoid分类器

CLC Number: