基于注意力机制的图像篡改检测

Image Tampering Detection Based on Attention Mechanism

目前已有的基于分割的图像篡改检测方法由于标注困难,可用的篡改数据集较少,造成训练数据的缺乏,同时篡改图像经过处理后边界难以识别,导致分割精度低。针对上述问题提出了基于注意力机制的图像篡改检测网络,该网络实现了篡改图像的生成,篡改区域的分割和优化。其中,生成器创建篡改图像用于扩充训练数据,基于注意力机制的分割优化模块用于增强篡改区域边界的特征提取能力,最后以实验结果证明了此方法的准确性和有效性。

At present, the existing image tamper detection methods based on segmentation are difficult to label and have few available tamper data sets, resulting in the lack of training data. At the same time, the boundary of the tampered image is difficult to identify after processing, resulting in low segmentation accuracy. To solve the above problems, an image tamper detection network based on attention mechanism is proposed. The network realizes the generation of tampered images, the segmentation and optimization of tampered regions. The generator creates the tampered image to expand the training data, and the segmentation optimization module based on attention mechanism is used to enhance the feature extraction ability of the tampered region boundary. Finally, the experimental results show the accuracy and effectiveness of this method.