Image Search System Using Segmented Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional search engine services are inadequate for identifying relevant information when users struggle to formulate effective search requests, especially for visual content like historical buildings or sculptures, and image-based search systems face challenges in fine-grain image matching due to computational expense.
Innovation Solution
An image search system that identifies visual parts of objects within a target image by analyzing multiple versions, using techniques like scale-invariant feature transform (SIFT) to filter out artifacts and rank similar images based on their likelihood of being duplicates, providing associated content as search results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If local features are used for fine-grain image matching, then object recognition accuracy is improved, but computational cost becomes too expensive to be practical
Solution Approach 1:
The patent segments the image matching process into two stages: first using CBIR with global image features to quickly identify candidate duplicate images, then using local features for fine-grain verification only on these candidates. This segmentation reduces computational cost by avoiding exhaustive local feature comparison across all images while maintaining high matching accuracy through the verification stage.
Solution Approach 2:
The patent applies partial action by using local features only partially - specifically, only on a subset of candidate images identified by CBIR rather than on all images in the database. This partial application of the computationally expensive local feature method maintains accuracy where needed while reducing overall computational burden.
2Productivity
If CBIR systems are used for image search, then search speed is improved, but fine-grain image matching capability is lost
Solution Approach 1:
The patent segments the image search process into two distinct phases: a first phase using CBIR for rapid candidate identification based on global image characteristics, and a second phase using local features for precise fine-grain matching on candidates only. This segmentation enables the system to achieve both fast search speeds through CBIR and high matching accuracy through subsequent local feature verification.
Solution Approach 2:
The patent introduces an intermediary step where CBIR results serve as a bridge between the initial image database and the final fine-grain matching stage. The CBIR system acts as a mediator that filters and prioritizes candidate images before they undergo computationally intensive local feature analysis, enabling efficient progression from speed-oriented search to accuracy-oriented verification.
3Measurement precision
If multiple image versions are analyzed, then duplicate detection accuracy is improved, but processing time increases
Solution Approach 1:
The patent segments the analysis of multiple image versions into hierarchical stages: first analyzing global features across all versions using CBIR to identify potential duplicates quickly, then performing detailed local feature analysis only on the most promising candidates. This segmented approach maintains high duplicate detection accuracy by considering multiple versions while minimizing processing time through selective detailed analysis.
Data Source
AI summary
A method and system for searching for content relating to a target or query image by identifying duplicate images with associated content is provided. An image search system identifies visual parts of objects within the target image based on analysis of two or more versions of the target image. The image search system identifies visual parts based on analysis of the versions. The image search system then identifies images of an image database that have visual parts that are similar to the visual parts of the target image. The image search system may rank the identified images based on their likelihood of being duplicates of the target image and provide their associated content as the search result ordered according to the ranking of the images.