Image Retrieval Using Multi-Viewpoint Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image retrieval methods for industrial design examination struggle to effectively compare and retrieve similar designs from diverse viewpoints, leading to suboptimal similarity determination and display of results.
Innovation Solution
A computer-based image retrieval method that reads and extracts feature quantities from multiple image groups stored in databases, calculates similarity using Euclidean distance, and displays results in order of similarity, allowing for effective comparison and retrieval of similar objects across different viewpoints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple image groups from different viewpoints are stored and compared, then retrieval accuracy is improved, but system complexity and processing time increase
Solution Approach 1:
The patent segments the image comparison task by extracting feature quantities from individual images within image groups separately, then combining results. Each image group is processed independently to extract features, and similarity is determined by aggregating these individual comparisons, making the complex multi-viewpoint comparison manageable and systematic
Solution Approach 2:
The patent transitions from comparing single images to comparing multiple image groups from different viewpoints by adding the dimension of viewpoint diversity. Instead of single-point comparison, the system evaluates similarity across multiple dimensional perspectives (different viewpoints), enriching the retrieval accuracy through multi-dimensional feature quantity comparison
2Measurement precision
If multiple image groups from different viewpoints are stored and compared, then retrieval accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies preliminary action by extracting and storing feature quantities from images in advance before the actual retrieval operation. The feature quantities are pre-processed and prepared, so that during retrieval, the system only needs to compare pre-extracted features rather than processing raw images from multiple viewpoints in real-time, significantly reducing processing time
Solution Approach 2:
The patent segments the processing workflow into distinct phases: feature extraction phase (performed preliminarily) and similarity determination phase (performed during retrieval). This segmentation allows computationally intensive feature extraction to be done in advance, separating it from the time-critical retrieval operation, thereby reducing actual retrieval processing time while maintaining accuracy
3Measurement precision
If image feature quantities from multiple viewpoints are extracted and compared, then similarity determination accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent extracts essential feature quantities from images, separating the critical similarity-determining characteristics from the complete image data. By taking out only the relevant feature quantities (such as geometric features, color features, or texture features) rather than processing entire images, the system reduces computational complexity while preserving the ability to accurately determine similarity across multiple viewpoints
Solution Approach 2:
The patent transforms images into parameter representations (feature quantities) that capture essential characteristics in a condensed form. This parameter transformation changes the data representation from complex image matrices to manageable feature vectors, enabling efficient comparison while maintaining accuracy in similarity determination across different viewpoints
Data Source
AI summary
An image retrieval method includes reading from a first database a first image group mapped to an object as a retrieval target, and extracting an image feature quantity of a plurality of images of the first image group. A second image group mapped to an object as a search key is read from a second database and an image feature quantity of a plurality of images of the second image group is extracted. Similarity is determined between the first image group and the second image group, based on the image feature quantity of the plurality of images of the first image group and the image feature quantity of the plurality of images of the second image group. On an output device, as retrieval results, the plurality of images of the first image group are displayed in order of similarity of the object as the retrieval target based on determined similarity.