《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (7): 2288-2294.DOI: 10.11772/j.issn.1001-9081.2022060840

• 多媒体计算与计算机仿真 • 上一篇    

基于主动判别机制的自适应生成对抗网络图像去模糊算法

刘安阳1, 赵怀慈2, 蔡文龙1(), 许泽超2, 解瑞灯1   

  1. 1.北华航天工业学院 电子与控制工程学院, 河北 廊坊 065000
    2.中国科学院 沈阳自动化研究所, 沈阳 110169
  • 收稿日期:2022-06-10 修回日期:2022-08-24 接受日期:2022-08-26 发布日期:2022-09-22 出版日期:2023-07-10
  • 通讯作者: 蔡文龙
  • 作者简介:刘安阳(1997—),男,山东济南人,硕士研究生,主要研究方向:人工智能、图像处理、模式识别;
    赵怀慈(1974—),男,山东潍坊人,研究员,博士,主要研究方向:图像处理、复杂系统建模与仿真、通信与信息处理;
    蔡文龙(1976—),男,河北衡水人,副教授,硕士,主要研究方向:智能控制、微弱信号处理;
    许泽超(1996—),男,山东德州人,硕士,主要研究方向:图像分割、深度学习;
    解瑞灯(1996—),男,山东菏泽人,硕士研究生,主要研究方向:控制工程、计算机工程自动化设计。
  • 基金资助:
    国家装备重大基础研究项目(51405-02A01);河北省重点研发计划项目(20327216D)

Adaptive image deblurring generative adversarial network algorithm based on active discrimination mechanism

Anyang LIU1, Huaici ZHAO2, Wenlong CAI1(), Zechao XU2, Ruideng XIE1   

  1. 1.School of Electronics and Control Engineering,North China Institute of Aerospace Engineering,Langfang Hebei 065000,China
    2.Shenyang Institute of Automation,Chinese Academy of Sciences,Shenyang Liaoning 110169,China
  • Received:2022-06-10 Revised:2022-08-24 Accepted:2022-08-26 Online:2022-09-22 Published:2023-07-10
  • Contact: Wenlong CAI
  • About author:LIU Anyang, born in 1997, M. S. candidate. His research interests include artificial intelligence, image processing, pattern recognition.
    ZHAO Huaici, born in 1974, Ph. D., research fellow. His research interests include image processing, complex system modeling and simulation, communication and information processing.
    CAI Wenlong, born in 1976, M. S., associate professor. His research interests include intelligent control, weak signal processing.
    XU Zechao, born in 1996, M. S. His research interests include image segmentation, deep learning.
    XIE Ruideng, born in 1996, M. S. candidate. His research interests include control engineering, automation design in computer engineering.
  • Supported by:
    Major Basic Research Project of National Equipment of China(51405-02A01);Key Research and Development Program of Hebei Province(20327216D)

摘要:

针对现有图像去模糊算法在处理边缘丢失时出现弥散和伪影以及在视频处理中使用全帧去模糊方式导致不满足实时性需求的问题,提出一种基于主动判别机制的自适应生成对抗网络图像去模糊(ADBGAN)算法。首先,提出一种自适应模糊判别机制,开发了自适应模糊处理网络模块对输入图像进行模糊先验判断。在采集到输入时提前判断输入图像的模糊程度,从而剔除足够清晰的输入帧以提升算法运行效率。然后,在精细特征提取过程中引入注意力机制中的激励环节,从而在特征提取的流程中进行权重归一化来提升网络对精细特征的恢复能力。最后,在生成器架构中改进了特征金字塔精细特征恢复结构,并采用更轻量化的特征融合流程提高运行效率。为验证算法的有效性,在开源数据集GoPro和Kohler上进行了详细的对比实验。实验结果显示,在GoPro数据集中ADBGAN的视觉保真度是尺度循环网络(SRN)算法的2.1倍,并在峰值信噪比(PSNR)上较SRN算法提升了0.762 dB,具有良好的图像信息恢复能力;在视频数据处理时间上ADBGAN大幅超越了测试的所有算法,实测处理时间较SRN减少了85.9%。ADBGAN能够高效生成信息质量更高的去模糊图像。

关键词: 图像去模糊, 图像去噪, 生成对抗网络, 自适应判别, 特征恢复

Abstract:

Aiming at the problems that existing image deblurring algorithms suffer from diffusion and artifacts when dealing with edge loss and the use of full-frame deblurring in video processing does not meet real-time requirements, an Adaptive DeBlurring Generative Adversarial Network (ADBGAN)algorithm based on active discrimination mechanism was proposed. Firstly, an adaptive fuzzy discrimination mechanism was proposed, and an adaptive fuzzy processing network module was developed to make a priori judgment of fuzziness on the input image. When collecting the input, the blurring degree of the input image was judged in advance, and the input frame which was clear enough was eliminated to improve the running efficiency of the algorithm. Then, the incentive link of the attention mechanism was introduced in the process of fine feature extraction, so that weight normalization was carried out in the forward flow of feature extraction to improve the performance of the network to recover fine-grained features. Finally, the feature pyramid fine feature recovery structure was improved in the generator architecture, and a more lightweight feature fusion process was adopted to improve the running efficiency. In order to verify the effectiveness of the algorithm, detailed comparison experiments were conducted on the open source datasets GoPro and Kohler. Experimental results on GoPro dataset show that the visual fidelity of ADBGAN is 2.1 times that of Scale-Recurrent Network (SRN) algorithm, the Peak Signal-to-Noise Ratio (PSNR) of ADBGAN is improved by 0.762 dB compared with that of SRN algorithm, and ADBGAN has good image information recovery ability; in terms of video processing time,the actual processing time is reduced by 85.9% compared to SRN.The proposed algorithm can generate deblurred images with higher information quality efficiently.

Key words: image deblurring, image denoising, Generative Adversarial Network (GAN), adaptive discrimination, feature recovery

中图分类号: