Related Experiment Video
Updated: Jun 23, 2026
11:49
Cereal Crop Ear Counting in Field Conditions Using Zenithal RGB Images
Published on: February 2, 2019
Source camera identification for heavily JPEG compressed low resolution still images.
Erwin J Alles1, Zeno J M H Geradts, Cor J Veenman
1Department of Digital Technology and Biometry, Netherlands Forensic Institute, The Hague, Holland.
Journal of Forensic Sciences
|May 13, 2009
Summary
This study shows Photo Response Non-Uniformity (PRNU) effectively identifies the source camera of compressed digital photos. Simultaneous identification of multiple images significantly improves accuracy and reduces errors.
Area of Science:
- Digital Image Forensics
- Sensor Pattern Noise Analysis
Background:
- Digital photographs contain unique sensor noise patterns.
- Photo Response Non-Uniformity (PRNU) is a specific type of fixed pattern noise.
- Identifying the source camera is crucial for digital evidence authentication.
Purpose of the Study:
- To evaluate the effectiveness of PRNU for source camera identification.
- To assess performance under heavy JPEG compression and low resolution (640x480).
- To analyze both closed-set and open-set identification scenarios.
Main Methods:
- PRNU patterns were extracted using a 2D Gaussian filter.
- Reference and questioned image patterns were compared using correlation coefficients.
- Evaluated performance on single and multiple simultaneously identified images.
Main Results:
- High accuracy achieved in closed-set identification: 83% for single images, 100% for ~20 images.
- Successful identification from a diverse set of 38 cameras.
- Open-set identification showed improved accuracy with multiple images, overcoming initial high false rejection rates for single images.
Conclusions:
- PRNU is a viable forensic technique for source camera identification, even with heavy JPEG compression.
- Simultaneous identification of multiple images significantly enhances forensic analysis robustness.
- The method shows promise for real-world forensic applications requiring high confidence in image source attribution.
