Image Fingerprinting via Local Feature Hashing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting and preventing unauthorized disclosure of secure image data are inefficient due to their inability to effectively handle derivative forms of images, such as format changes, cropping, and rotations, and they suffer from high computational costs and storage requirements for large descriptors.
Innovation Solution
The method involves normalizing image data, identifying feature points, determining anchor points, calculating distances and angles relative to these points, and using a hash function to generate a fingerprint that is robust against variations, allowing for efficient storage and matching of secure image data across the organization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If global image features (entire image histograms) are used for detection, then detection coverage is improved, but the method fails to detect partial image matches and transformations
Solution Approach 1:
The patent divides the image into multiple local regions and extracts features from each region independently. Instead of using global histograms, the method creates local histograms for each image block, allowing partial matches to be detected while maintaining robustness against transformations through feature aggregation across regions.
2Adaptability or versatility
If local region features are used for detection, then ability to detect derivative forms is improved, but detection coverage for full image matches decreases
Solution Approach 1:
The patent merges local region features by aggregating local histograms into a global fingerprint representation. This combination allows the system to detect both partial matches (through local region independence) and full image matches (through aggregated global representation), resolving the contradiction between local and global detection capabilities.
3Measurement precision
If detailed descriptors are computed for feature points, then detection precision is improved, but storage requirements and computational cost increase
Solution Approach 1:
The patent extracts only the essential features needed for detection by computing histograms of color and texture properties rather than storing complete image data or complex descriptors. This extraction approach maintains detection precision while significantly reducing storage requirements and computational overhead.
4Measurement precision
If complete image data is stored for comparison, then detection accuracy is improved, but storage costs and processing time increase
Solution Approach 1:
The patent creates simplified fingerprint copies of images using histogram-based representations instead of storing complete image data. These fingerprint copies retain the essential characteristics needed for accurate detection while enabling rapid comparison operations, thus reducing processing time without sacrificing detection accuracy.
Data Source
AI summary
The present invention provides methods and systems to protect an organization's secure image information from unauthorized disclosure. In one embodiment, methods and systems to generate image fingerprints are provided. The fingerprints are generated for each feature point of an image data. Because the fingerprints take into account the neighborhood features around each feature point, the image fingerprints are robust against derivative images where the original image may have been altered. Methods and systems to maintain a fingerprint database for an organization's secure image data is also provided. In one embodiment, client fingerprints are generated for image data that a user intends to transmit outside of the organization. In some embodiments, suitable security actions are initiated if any of the client fingerprints match any of the fingerprints in the fingerprint database.