Image Retrieval Apparatus with Category-Based Additional Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image retrieval methods face challenges in efficiently finding specific images due to oversight in retrieval results, particularly when high-importance categories have few images, leading to reduced user convenience, and adjusting image recognition settings to minimize oversight can result in excessively large retrieval results.
Innovation Solution
An image retrieval apparatus with an image obtaining unit, basic searching unit, counting unit, determining unit, and additional searching unit that performs a detailed search in categories with insufficient specific images, ensuring that even partially visible or smaller images are included in the results, thereby reducing oversight and improving user convenience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If image recognition technique is used to retrieve specific images, then retrieval convenience is improved, but oversight occurs in retrieval results due to image misrecognition or recognition settings
Solution Approach 1:
The retrieval process is divided into two stages: basic search using image recognition technique and additional search for categories with insufficient results. This segmentation allows the system to leverage the convenience of automated recognition while addressing its reliability issues through a supplementary manual verification process for critical categories.
Solution Approach 2:
The system performs preliminary classification of images into categories based on recognition results, then identifies categories with insufficient specific images before conducting additional searches. This preliminary action enables targeted improvement of retrieval accuracy in problematic areas without affecting the overall efficient retrieval process.
2Reliability
If image recognition setting threshold is adjusted to reduce oversight, then retrieval accuracy is improved, but the number of images shown becomes too large, increasing user troublesomeness
Solution Approach 1:
The system applies different search strategies to different categories based on their specific characteristics. For categories with sufficient specific images, standard recognition thresholds are used to maintain efficiency. For categories with insufficient images, additional search is performed to ensure completeness. This local differentiation allows the system to optimize both accuracy and convenience without compromising either aspect globally.
3Reliability
If additional search is performed for all categories, then oversight is reduced, but the number of images to review increases, worsening user convenience
Solution Approach 1:
The search process is made dynamic by adjusting the search depth based on category-specific conditions. The system automatically determines whether additional search is needed for each category based on the count of specific images found in the basic search. This dynamic approach ensures comprehensive retrieval accuracy where needed while maintaining simple, efficient processing where standard search suffices.
Data Source
AI summary
An image retrieving apparatus is equipped with: an image obtaining unit for obtaining a plurality of images; a category obtaining unit for obtaining a category to which each of the obtained images belongs; a basic searching unit for searching a specific image including a specific object, from the obtained images; a counting unit for counting the number of the specific images for each category obtained; a determining unit for determining whether or not an additional search for the specific image is necessary for each category, based on the counted number of the specific images of each category; and an additional searching unit for searching the specific image from among the images which belong to the category for which it has been determined that the additional search is necessary and from which the searched specific image has been excluded.