Object-Based Visual Search via Region Segmentation and Feature Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search techniques, such as keyword searching, are inefficient in discovering relevant digital content due to the complexity of finding visually similar objects and their stylistic combinations within images.
Innovation Solution
A system and method that allows users to select objects of interest from images, generating feature vectors for these objects, which are then compared to stored feature vectors to retrieve visually similar images, including how these objects are combined with other objects, and can supplement text-based searches with visual refinements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If keyword searching is used to search digital content, then the search process is simple to implement, but the search efficiency and accuracy deteriorate due to the ever-expanding amount of accessible digital content
Solution Approach 1:
The patent segments images into multiple regions and generates separate feature vectors for each region. This allows the search system to compare specific object regions rather than entire images, significantly improving search efficiency and accuracy while maintaining manageable complexity through automated processing.
Solution Approach 2:
The patent introduces feature vectors as an intermediary representation between images and search queries. By converting images into feature vectors and comparing these representations, the system achieves efficient and accurate content-based search without requiring complex keyword matching or manual tagging.
2Reliability
If entire images are compared for visual similarity search, then comprehensive image matching is achieved, but the search precision deteriorates because the search cannot focus on specific objects within images
Solution Approach 1:
The patent divides images into multiple regions and generates feature vectors for each region separately. This segmentation enables precise object-level comparison while maintaining comprehensive image matching capability, as the system can identify and compare specific objects of interest within the broader image context.
Solution Approach 2:
The patent applies different processing and comparison focus to different regions of images. By generating region-specific feature vectors, the system can prioritize comparison of relevant object regions while still considering the overall image context, thereby achieving both precision and comprehensiveness.
3Measurement precision
If region-based feature vector comparison is used to improve search precision, then object-level search accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent performs preliminary image segmentation and feature vector generation for all images in the database before search execution. This pre-processing creates ready-to-compare feature representations, significantly reducing the computational complexity during actual search operations while maintaining high precision object-level comparison.
Data Source
AI summary
Described is a system and method for enabling visual search for information. With each selection of an object included in an image, additional images that include visually similar objects are determined and presented to the user.