Image Editing Feature Clustering for Guided Image Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image search technologies classify images based on data generation conditions rather than image features, making it difficult to provide effective editing support for users selecting images for design creation.
Innovation Solution
An image editing device that extracts image feature values, classifies images based on these values using cluster analysis, and provides editing support information by distinguishing images into sets based on their similarity and prevalence, including 'typical', 'novel', and 'peculiar' categories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If images are classified based on data generation conditions, then images can be organized systematically, but it becomes difficult to provide effective editing support for users selecting images for design creation
Solution Approach 1:
The patent changes the classification parameters from data generation conditions to image feature values (color, shape, texture, etc.). This allows images to be automatically classified while providing meaningful editing support, as feature-based classification directly reflects visual characteristics that users need for design work.
Solution Approach 2:
The patent segments images into multiple clusters based on feature values, creating distinct groups that represent different visual characteristics. This segmentation enables users to easily navigate and select images with specific features, improving ease of operation while maintaining systematic organization.
2Loss of information
If all searched images are displayed to users, then complete search results are provided, but users find it difficult to identify images with desired impressions among many images
Solution Approach 1:
The patent segments the complete set of searched images into multiple clusters based on feature values. Each cluster represents a distinct visual category, allowing users to quickly identify images with desired impressions without losing access to the complete search results. The segmentation is visually presented to guide user selection.
Solution Approach 2:
The patent applies different display characteristics to different clusters, with each cluster being visually distinguished to highlight its unique features. This local quality approach helps users quickly identify images with specific impressions by presenting clustered results with enhanced visual cues.
3Measurement precision
If images are classified into multiple fine-grained categories, then precise image features are captured, but the complexity of the classification system increases
Solution Approach 1:
The patent uses a small number of key image feature values (color, shape, texture) to achieve precise classification. By focusing on these fundamental parameters rather than attempting to classify across all possible image attributes, the system maintains measurement precision while avoiding excessive complexity.
Solution Approach 2:
The patent applies classification to only the most essential image features rather than all possible attributes. This partial action approach captures the most important visual characteristics needed for design work while keeping the classification system manageable and computationally efficient.
Data Source
AI summary
An image editing device includes: an extractor that extracts an image feature value from a plurality of images obtained by search; a classifier that classifies the plurality of images based on the extracted image feature value; and a provider that provides editing support information by causing a display to display a classification result by the classifier. The classifier sets, as a first image set, images classified as belonging to a cluster including a predetermined ratio or more of the plurality of images with respect to a total number of the plurality of images by classification based on the image feature value. The provider causes the display to display the images belonging to the first image set and other images distinguishably.