Journal of Computer Applications ›› 2023, Vol. 43 ›› Issue (9): 2692-2699.DOI: 10.11772/j.issn.1001-9081.2022091405

• 2022 10th CCF Conference on Big Data • Previous Articles     Next Articles

Lightweight image tamper localization algorithm based on large kernel attention convolution

Hong WANG, Qing QIAN(), Huan WANG, Yong LONG   

  1. School of Information,Guizhou University of Finance and Economics,Guiyang Guizhou 550025,China
  • Received:2022-09-19 Revised:2022-10-18 Accepted:2022-10-21 Online:2023-09-10 Published:2023-09-10
  • Contact: Qing QIAN
  • About author:WANG Hong, born in 1995, M. S. candidate. His research interests include artificial intelligence, passive image forensics.
    WANG Huan, born in 1987, Ph. D., lecturer. Her research interests include image forensics, multimedia security.
    LONG Yong, born in 1997, M. S. candidate. Her research interests include multimedia forensics, image steganography.
  • Supported by:
    National Natural Science Foundation of China(61902085);Science and Technology Program of Guizhou Province (QianKeHeJiChu-ZK[2021]General 311), Natural Science Research Project of Department of Education of Guizhou Province (QianJiaoHe KY [2021]136)


王宏, 钱清(), 王欢, 龙永   

  1. 贵州财经大学 信息学院,贵阳 550025
  • 通讯作者: 钱清
  • 作者简介:王宏(1995—),男,四川南充人,硕士研究生,CCF会员,主要研究方向:人工智能、图像被动取证
  • 基金资助:


Convolutional Neural Networks (CNN) are used for image forensics because of their high recognizable property, easy understanding, and strong learnability. However, their inherent disadvantages of the receptive field increasing slowly and neglecting long-range dependencies, and high computational cost cause the unsatisfactory accuracy and lightweight deployment of deep learning algorithms. To solve the above problems, a lightweight network-based image copy-paste tamper detection algorithm namely LKA-EfficientNet (Large Kernel Attention EfficientNet) was proposed. The characteristics of long-range dependencies and global receptive field were contained in LKA-EfficientNet, and the number of EfficientNetV2 parameters was optimized. As a result, the localization speed and detection accuracy of image tamper were improved. Firstly, the image was inputted into and processed in the backbone network based on Large Kernel Attention (LKA) to obtain the candidate feature maps. Then, the feature maps of different scales were used to construct the feature pyramid for feature matching. Finally, the candidate feature maps after feature matching were fused to locate the tampered area of the image. In addition, the triple cross entropy loss function was used by LKA-EfficientNet to further improve the accuracy of the algorithm in image tamper localization. Experimental results show that LKA-EfficientNet can not only reduce the floating-point operations by 29.54% but also increase the F1 by 4.88% compared to the same type algorithm — Dense-InceptionNet. The above verifies that LKA-EfficientNet can reduce computational cost and maintain high detection performance at the same time.

Key words: image tamper detection, lightweight network, attention mechanism, multi-scale feature pyramid, passive forensics


卷积神经网络(CNN)因辨识度高、易于理解、可学习性强而被用于图像取证,但它固有的感受野增加缓慢、忽略长端依赖性、计算量庞大等缺点导致深度学习算法的精度与轻量化部署效果并不理想,不适用于以轻量化形式实现图像篡改定位的场景。为解决上述问题,提出一种基于轻量化网络的图像复制-粘贴篡改检测算法——LKA-EfficientNet(Large Kernel Attention EfficientNet)。LKA-EfficientNet具有长端依赖性和全局感受野的特性,且优化了EfficientNetV2的参数量,提高了图像篡改定位速度和精度。首先,将输入图像通过基于大核注意力(LKA)卷积的基干网络进行处理,得到候选特征图;随后,使用不同尺寸的特征图构建特征金字塔进行特征匹配;最后,将特征匹配后的特征图进行融合以定位图像篡改区域;此外,LKA-EfficientNet使用三元组交叉熵损失函数进一步提升了算法定位篡改图像的精度。实验结果表明,LKA-EfficientNet与同类型的Dense-InceptionNet算法相比,不仅能够降低29.54%的浮点运算量,而且F1分数也提高了4.88%,验证了LKA-EfficientNet可以在保持高检测性能的同时降低计算量。

关键词: 图像篡改检测, 轻量化网络, 注意力机制, 多尺度特征金字塔, 被动取证

CLC Number: