Image Matching Using Feature Vectors for Distorted Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image protection schemes are ineffective in detecting unauthorized uses of images that have been cropped, resized, rotated, or otherwise distorted, and struggle to identify redigitized versions, especially when images are scanned from printed originals.
Innovation Solution
A system and method for image matching that uses an image processing module to extract principal features and construct multi-dimensional vectors, allowing for resilient comparisons between query and catalog images despite distortions, and includes a filtering module to select candidate images based on distance measures and transformations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional image protection schemes are used to detect unauthorized uses, then direct copies can be identified, but distorted versions (cropped, resized, rotated, scanned) cannot be detected
Solution Approach 1:
The image is divided into multiple blocks or regions, and feature points are extracted from each block. This segmentation allows the system to identify characteristic patterns that remain recognizable even when the image is distorted, cropped, or transformed, thereby improving detection reliability across various distortion types.
Solution Approach 2:
The system transforms image data into different parameter spaces (e.g., frequency domain, feature point coordinates, descriptor vectors) that are more robust to distortion. By changing the representation parameters rather than working with raw pixel data, the system can identify distorted versions while maintaining detection reliability.
2Measurement precision
If comprehensive image comparison is performed on large image sets, then accurate matches can be identified, but the number of searches and processing time increase significantly
Solution Approach 1:
Feature points and descriptors are pre-computed and stored for all images in the database before actual comparison is needed. This preliminary extraction of characteristic features allows rapid comparison during query operations, maintaining high match accuracy while dramatically improving search efficiency for large image sets.
Solution Approach 2:
The system extracts only the most salient feature points and descriptors from images, separating the essential identification information from the complete image data. This extraction enables efficient comparison by working with compact feature representations rather than full-resolution images, thus improving productivity without sacrificing measurement precision.
3Productivity
If feature-based image matching is used to reduce search complexity, then processing speed improves, but effectiveness in identifying redigitized and heavily distorted images decreases
Solution Approach 1:
The system moves from two-dimensional spatial feature matching to multi-dimensional feature space comparison by extracting descriptors that capture contextual relationships, gradient orientations, and scale information. This dimensional transformation enables the matching algorithm to identify distorted images more reliably while maintaining processing speed through efficient vector comparisons.
Data Source
AI summary
A system and method to detect similarities between images. The system and method allow comparisons between a query image and one or more catalog images in a manner that is resilient to scanning, scaling, rotating, cropping and other distortions of the query image. The system includes an image processing module that determines and/or calculates principle features of a catalog image and constructs a feature vector using one or more of the principle features. The system also includes a matching module that matches a query image to one or more catalog images. The system finds matches based on a distance measure of features present in the query image and features present in the catalog images.