Image Splicing Tamper Detection Based on Deep Learning and Attention Mechanism
Cheng‐Xiang Wang, Yan Li, Guangli Wu · 2021 IEEE 6th International Conference on Signal and Image Processing (ICSIP) · 2021
Aiming at the low detection accuracy and poor positioning effect of traditional image Splicing tampering detection methods, an image Splicing tampering detection method based on deep learning and attention mechanism was proposed. Firstly, by convolution neural network (CNN) to extract the image features, the features extracted from CNN were input into the channel attention mechanism module to generate attention characteristic figure; Then, the convolution of 1*1 is used to replace the original full connection layer in the classification, realizing tampering image classification. Finally, the method of adversarial complementary learning (ACOL) is used to automatically locate the tamper area of semantic interest, which is mapped to the original image in the form of heat map to locate the tampering area. The experimental results show that, on CASIA1.0 and CASIA2.0 data sets, the algorithm proposed in this paper can effectively improve the accuracy of Mosaic image tamper detection and locate the tamper area. As for Accuracy, the model in this paper is superior to the existing image Mosaic tamper detection algorithm, and the prediction Accuracy of the two data sets reaches 99.60% and 99.92% respectively.