Representative Image Search Using Clustering and Region Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users face difficulties in accurately and efficiently searching large collections of images based on fragmentary memories, as existing methods require precise illustration of search images or selection from predefined parts that may not match actual image components.
Innovation Solution
An image search apparatus and method that extracts representative images, allows users to set search conditions using parts or all of these images, including position, and enables region extraction and image processing to facilitate easier and more accurate searches based on fragmentary memories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If users set search conditions by illustrating entire images, then search accuracy is improved, but ease of operation deteriorates because users find illustrating difficult and time-consuming
Solution Approach 1:
The patent segments the image search process into two stages: first, the system automatically extracts and clusters similar images to generate representative images; second, users set search conditions by selecting from these pre-generated representative images rather than illustrating from scratch. This segmentation reduces user effort while maintaining search accuracy.
Solution Approach 2:
The system performs preliminary action by automatically generating representative images through clustering before the user sets search conditions. These representative images serve as pre-prepared search templates that users can directly select or modify, eliminating the need for users to perform the time-consuming illustration task.
2Ease of operation
If users set search conditions by selecting from prepared image parts, then ease of operation is improved, but search accuracy deteriorates because the prepared parts may not fit well to remembered parts in actual images
Solution Approach 1:
The system enables self-service by automatically generating representative images through clustering algorithms that analyze the actual image database. These self-generated representative images inherently reflect the actual content and structure of images in the database, ensuring they fit well with remembered parts without requiring manual preparation.
Solution Approach 2:
The system changes the parameter of search condition representation from manually prepared generic image parts to dynamically generated representative images that adapt to the specific content of the image database. This parameter change allows the search conditions to better match actual image content while maintaining ease of selection.
3Quantity of substance
If the system processes large numbers of images, then completeness of image collection is improved, but loss of time increases due to the tedious classification task
Solution Approach 1:
The patent replaces the mechanical manual classification process with an automated computational system that performs image clustering and representative image extraction. This substitution eliminates the time-consuming manual sorting and classification task while enabling processing of large image collections.
Solution Approach 2:
The system changes the approach from manual classification to automatic clustering based on image similarity metrics. This parameter change in the classification method dramatically reduces the time required to process and organize large numbers of images while maintaining completeness.
Data Source
AI summary
One or more representative images extracted from an image group comprising a plurality of images is/are displayed. A part or all of the representative image or images, such as a main subject region or a background region including a search target, is/are selected from the representative image or images, and used for setting search conditions. The image group is searched for an image or images agreeing with the search conditions having been set.