Digital Image Anonymization via PRNU Noise Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for eliminating PRNU noise from digital images are either ineffective, require specialized equipment and expertise, or necessitate physical access to the source device, limiting their practicality and ability to completely anonymize images.
Innovation Solution
A system and method that estimates and eliminates PRNU noise by using a test image and reference images captured with the same sensor, calculating the PRNU noise power factor through correlation, and subtracting the estimated noise from the test image, without requiring physical access or special shooting conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If flat-fielding method is used to weaken sensor noise, then temperature-dependent sensor noise is reduced, but PRNU noise is only slightly weakened and physical access to source device is required
Solution Approach 1:
The patent extracts PRNU noise from digital images using statistical analysis and signal processing techniques. By separating the PRNU component from the image data, the method eliminates the need for physical access to the source device while effectively removing the harmful noise fingerprint that enables camera identification.
Solution Approach 2:
The patent replaces the mechanical/physical flat-fielding process with a digital signal processing approach. Instead of requiring physical capture of dark and flat frames using specialized equipment, the method uses computational algorithms to estimate and remove PRNU noise from existing images, substituting mechanical operations with electronic processing.
2Object-affected harmful factors
If multiple captures with same camera are used to remove temperature-based sensor noise, then noise removal is achieved, but expertise and physical access are required
Solution Approach 1:
The patent enables self-service PRNU removal by allowing users to process their own images without requiring expertise in noise analysis or physical access to camera devices. The automated algorithm performs statistical analysis and noise extraction independently, making the process accessible to ordinary users while effectively removing temperature-based sensor noise.
3Measurement precision
If PRNU noise is used for source device identification, then camera fingerprinting is achieved, but privacy is hindered
Solution Approach 1:
The patent converts the harmful PRNU noise that enables privacy violations into a removable artifact. By developing methods to extract and eliminate PRNU fingerprints from images, the technology transforms the previously harmful identification feature into a removable element, thereby protecting privacy while maintaining the ability to analyze noise characteristics for legitimate purposes.
4Object-affected harmful factors
If RAW outputs are required for flat-fielding, then noise removal effectiveness is improved, but applicability to consumer cameras is reduced
Solution Approach 1:
The patent creates a universal PRNU removal method that works across multiple camera formats and image types. The statistical analysis approach can process various image formats including JPEG and other compressed formats, not limited to RAW outputs. This multi-functional capability allows the method to be applied to consumer cameras, professional equipment, and different image types uniformly.
Data Source
AI summary
Disclosed is a method of anonymization of digital images through elimination of the Photo-Response Non Uniformity noise pattern which is unique to the imaging sensor and latent in all digital images taken by digital cameras or devices with imaging sensors.