Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PDC-ViT : Source Camera Identification using Pixel Difference Convolution and Vision Transformer

Omar Elharrouss, Younes Akbari, Noor Almaadeed, Somaya Al-Maadeed, Fouad Khelifi, Ahmed Bouridane

TL;DR

This work tackles source camera identification for images and videos by introducing PDC-ViT, a two-stage framework that first extracts discriminative pixel-difference features via Angular and Radial Pixel Difference Convolution (APDC and RPDC) and then classifies with a Vision Transformer. By feeding PDC features into a ViT, the method captures both local pixel-level variations and global context, achieving high accuracy across multiple public datasets. Empirical results show state-of-the-art performance on Vision, Daxing, Socrates, QUFVD, and Video-ACID, with notable improvements over strong baselines and robust behavior across diverse camera brands and models. The approach offers practical forensic benefits, including improved traceability of image provenance and robustness to content variations, while highlighting future directions in data efficiency, training stability, and handling emerging devices.

Abstract

Source camera identification has emerged as a vital solution to unlock incidents involving critical cases like terrorism, violence, and other criminal activities. The ability to trace the origin of an image/video can aid law enforcement agencies in gathering evidence and constructing the timeline of events. Moreover, identifying the owner of a certain device narrows down the area of search in a criminal investigation where smartphone devices are involved. This paper proposes a new pixel-based method for source camera identification, integrating Pixel Difference Convolution (PDC) with a Vision Transformer network (ViT), and named PDC-ViT. While the PDC acts as the backbone for feature extraction by exploiting Angular PDC (APDC) and Radial PDC (RPDC). These techniques enhance the capability to capture subtle variations in pixel information, which are crucial for distinguishing between different source cameras. The second part of the methodology focuses on classification, which is based on a Vision Transformer network. Unlike traditional methods that utilize image patches directly for training the classification network, the proposed approach uniquely inputs PDC features into the Vision Transformer network. To demonstrate the effectiveness of the PDC-ViT approach, it has been assessed on five different datasets, which include various image contents and video scenes. The method has also been compared with state-of-the-art source camera identification methods. Experimental results demonstrate the effectiveness and superiority of the proposed system in terms of accuracy and robustness when compared to its competitors. For example, our proposed PDC-ViT has achieved an accuracy of 94.30%, 84%, 94.22% and 92.29% using the Vision dataset, Daxing dataset, Socrates dataset and QUFVD dataset, respectively.

PDC-ViT : Source Camera Identification using Pixel Difference Convolution and Vision Transformer

TL;DR

This work tackles source camera identification for images and videos by introducing PDC-ViT, a two-stage framework that first extracts discriminative pixel-difference features via Angular and Radial Pixel Difference Convolution (APDC and RPDC) and then classifies with a Vision Transformer. By feeding PDC features into a ViT, the method captures both local pixel-level variations and global context, achieving high accuracy across multiple public datasets. Empirical results show state-of-the-art performance on Vision, Daxing, Socrates, QUFVD, and Video-ACID, with notable improvements over strong baselines and robust behavior across diverse camera brands and models. The approach offers practical forensic benefits, including improved traceability of image provenance and robustness to content variations, while highlighting future directions in data efficiency, training stability, and handling emerging devices.

Abstract

Source camera identification has emerged as a vital solution to unlock incidents involving critical cases like terrorism, violence, and other criminal activities. The ability to trace the origin of an image/video can aid law enforcement agencies in gathering evidence and constructing the timeline of events. Moreover, identifying the owner of a certain device narrows down the area of search in a criminal investigation where smartphone devices are involved. This paper proposes a new pixel-based method for source camera identification, integrating Pixel Difference Convolution (PDC) with a Vision Transformer network (ViT), and named PDC-ViT. While the PDC acts as the backbone for feature extraction by exploiting Angular PDC (APDC) and Radial PDC (RPDC). These techniques enhance the capability to capture subtle variations in pixel information, which are crucial for distinguishing between different source cameras. The second part of the methodology focuses on classification, which is based on a Vision Transformer network. Unlike traditional methods that utilize image patches directly for training the classification network, the proposed approach uniquely inputs PDC features into the Vision Transformer network. To demonstrate the effectiveness of the PDC-ViT approach, it has been assessed on five different datasets, which include various image contents and video scenes. The method has also been compared with state-of-the-art source camera identification methods. Experimental results demonstrate the effectiveness and superiority of the proposed system in terms of accuracy and robustness when compared to its competitors. For example, our proposed PDC-ViT has achieved an accuracy of 94.30%, 84%, 94.22% and 92.29% using the Vision dataset, Daxing dataset, Socrates dataset and QUFVD dataset, respectively.

Paper Structure

This paper contains 16 sections, 3 equations, 7 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. Proposed system
  4. Pixel Difference Convolution Backbone
  5. Proposed PDC-ViT Model
  6. Model Advantages and applications
  7. Experimental results
  8. Implementation setup
  9. Datasets
  10. Evaluation using the Vision dataset
  11. Evaluation using the Daxing dataset
  12. Evaluation using the Socrates dataset
  13. Evaluation using the QUFVD dataset
  14. Evaluation using the Video-ACID +Video WA PRNU dataset
  15. Challenges and future directions
  16. ...and 1 more sections

Figures (7)

  • Figure 1: Flowchart of the proposed network.
  • Figure 2: Selection of pixel pairs and convolution in Angular PDC.
  • Figure 3: Confusion matrices obtained of the evaluation using PDC-ViT on Vision and Daxing datasets.
  • Figure 4: t-SNE visualization on QUFVD and Vision dataset
  • Figure 5: Confusion matrices obtained of the evaluation using APDC-ViT, RPDC-ViT, and PDC-ViT (ARPDC-ViT ) proposed architectures on QUFVD dataset.
  • ...and 2 more figures