摘要
目的 恶意的图像拼接篡改给名誉、法律、政治等带来一系列的挑战,而现有的图像拼接检测算法通常采用参数固定的高通滤波器提取滤波特征进行预处理,没有考虑图像之间的差异。方法 本文设计自适应残差模块(adaptive residuals module, ARM)凸显拼接篡改痕迹,将卷积运算后的残差多次拼接,且每次拼接后再利用注意力机制实现通道间的非线性交互。然后,使用通道注意力SE(squeeze and excitation)模块以减少由ARM提取残差特征产生的通道之间信息冗余,并以在图像分类领域获得卓越性能的EfficientNet(high-efficiency network)为骨干网络,提出一种新的图像拼接检测算法。结果 实验结果表明,所提算法在CASIA I(CASIA image tampering detection evaluation database),CASIA II,COLUMBIA COLOR,NIST16(NIST special database 16)和FaceForensic++这5个公开数据集上分别取得98.95%,98.88%,100%,100%,88.20%的检测准确率,获得比现有算法更高的准确率。提出的ARM将骨干网络EfficientNet在CASIA II数据集的准确率提高了3.94%以上。结论 提出的基于自适应残差的图像拼接检测算法充分考虑图像之间的差异,凸显篡改区域与未篡改区域之间的区别,并获得更好的拼接检测结果。
Objective In recent years,digital media have become central to the exchange of information in our daily lives.With the rapid development of image editing tools and deep learning techniques,tampering with transmitted images iseasy.Image splicing is one of the most common types of image tampering.Malicious image splicing challenges reputation,law,and politics.Therefore,various approaches have been proposed for detecting image splicing forgeries.Deep learninghas also been successfully applied in image splicing detection.However,the existing deep learning-based works usuallypreprocess the input images by extracting features filtered by the high-pass filters with fixed parameters,which does notconsider the differences between images.Method Therefore,a new image splicing detection algorithm is proposed in thispaper.First,an adaptive residual module(ARM)is designed to highlight the splicing traces.In the ARM,the residual after the convolution operation is serialized several times,and the attention mechanism is used to realize the nonlinear inter⁃action between channels after each connection.Unlike ordinary filters with fixed parameters,the ARM module entirelyrelies on the feature reuse and attention mechanism of residuals to retain and enlarge the details of the splicing.Then,asqueeze and excitation(SE)module is used to reduce the inter channel information redundancy generated by ARM residualfeature extraction.The SE module uses an average adaptive pool to generate channel statistics information on global spaceand the gating mechanism of the Sigmoid activation function to learn channel weights from channel dependencies.Finally,a new image splicing detection algorithm is proposed by combining with the proposed ARM and the backbone network Eff⁃cientNet,a model with excellent performance in image classification.Result Experimental results show the proposed algo⁃rithm achieves 98.95%,98.88%,100%,100%,and 88.20%detection accuracies on CASIA image tampering detectionevaluation database(CASIA I),CASIA II,COLUMBIA COLOR,N
作者
张玲
穆文鹏
陈北京
Zhang Ling;Mu Wenpeng;Chen Beijing(Engineering Research Center of Digital Forensics,Ministry of Education,Nanjing University of Information Science&Technology,Nanjing 210044,China;Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology(CICAEET),Nanjing University of Information Science&Technology,Nanjing 210044,China;School of Computer Science,Nanjing University of Information Science&Technology,Nanjing 210044,China)
出处
《中国图象图形学报》
CSCD
北大核心
2024年第2期419-429,共11页
Journal of Image and Graphics
基金
国家自然科学基金项目(62072251)
江苏高校优势学科建设工程资助项目。
