Visual Search Characteristic Point Ambiguity Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing visual search algorithms in image databases face inefficiencies due to the high number of characteristic points detected, leading to increased computational complexity and resource wastage, particularly with repetitive patterns, which results in ambiguous point detection and insufficient rejection of non-matching points.
Innovation Solution
Implementing a method to classify characteristic points based on an ambiguity threshold, removing ambiguous points, and using spatial transformation models to link relevant points, thereby reducing unnecessary calculations and focusing on non-ambiguous points for comparison and relevance rating.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If characteristic point detection is performed using SIFT or SURF algorithms, then visual search capability is improved, but the number of characteristic points becomes excessively high, increasing computational complexity
Solution Approach 1:
The patent extracts and removes ambiguous characteristic points from the set of detected points before performing visual search comparisons. This extraction of problematic elements reduces the number of points to be processed, thereby decreasing computational complexity while preserving the accuracy of non-ambiguous point matching
Solution Approach 2:
The patent performs preliminary classification of characteristic points into ambiguous and non-ambiguous categories before the main visual search process. This preliminary action filters out points that would cause computational waste, allowing the subsequent matching process to operate more efficiently on a reduced set of reliable points
2Reliability
If all detected characteristic points are used for comparison, then comprehensive matching is achieved, but resource wastage increases due to processing ambiguous points from repetitive patterns
Solution Approach 1:
The patent extracts ambiguous characteristic points that arise from repetitive patterns and removes them from the comparison process. This selective extraction eliminates computational resource wastage on points that cannot provide reliable matching information, while maintaining matching reliability through careful classification
Solution Approach 2:
The patent applies different processing quality to different characteristic points based on their ambiguity status. Non-ambiguous points undergo full comparison processing while ambiguous points are filtered out, creating a local quality differentiation that optimizes resource allocation according to each point's reliability
3Measurement precision
If geometrical verification is performed on all similar points, then accurate object matching is achieved, but the number of affine transformation models to be tested increases significantly
Solution Approach 1:
The patent extracts and removes ambiguous characteristic points before performing geometrical verification. This reduction in the number of points to be verified decreases the number of affine transformation models that need to be tested, thereby improving search speed while maintaining matching accuracy through selective point processing
Data Source
AI summary
A method for extracting characteristic points from an image, includes extracting characteristic points from a first image, generating for each characteristic point a descriptor with several components describing an image region around the characteristic point, and comparing two by two the descriptors of the first image, the characteristic points whose descriptors have a proximity between them greater than an ambiguity threshold, being considered ambiguous.