ResNet-Based Camera Model Identification with Adaptive Preprocessing Module and Weight Fusion of Global Information
Boru Chen, Waleed Habib Abdulla · 2023
Identifying camera models used in digital images is an essential component of image forensics, which plays a vital role in preventing image-related crimes. While deep learning techniques, especially Convolutional Neural Networks (CNNs), are currently the most commonly used approach for Camera Model Identification (CMI) due to their high efficiency and accuracy, the performance of the CNN-based method will be adversely affected by the image content due to its structure.In this study, we introduce a novel approach to CMI, which utilizes the Residual Neural Network (ResNet), Adaptive Preprocessing Module (APM), and Weight Fusion of Global Information (WFGI) to overcome this issue. The results of the experiment indicate that our proposed method has achieved an accuracy exceeding 97.5% when tested on a large-scale dataset that consists of 45 different camera models based on the Vision and Dresden database. Furthermore, our approach exhibits robustness under different attack scenarios, thereby highlighting its potential for practical use in image forensics applications.