Dynamic Image Clustering Through Iterative Group Division
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image clustering methods achieve insufficient clustering degrees, as the number of classes is predetermined, limiting the aggregation and separation of images within and between groups.
Innovation Solution
An image clustering method that determines a first clustering parameter for M image groups, divides each group into two to obtain M+1 reference groups, and selects the highest second clustering parameter to further divide groups when necessary, improving the clustering degree by iteratively refining the group divisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the number of classes is predetermined, then the clustering process is simple and efficient, but the clustering degree is insufficient
Solution Approach 1:
The patent transforms the static predetermined number of classes into a dynamic structure where classes can be automatically adjusted. The clustering algorithm iteratively divides image groups based on clustering parameters, allowing the number of classes to evolve from an initial guess to an optimal value determined by the data itself, thus resolving the contradiction between simplicity and precision
Solution Approach 2:
The patent changes the parameter of class number from a fixed input to a dynamically determined output. By using clustering parameters (such as silhouette coefficient or Davies-Bouldin index) to evaluate and guide the division process, the system automatically adjusts the number of classes to achieve optimal clustering degree while maintaining computational efficiency
2Measurement precision
If images are divided into more groups, then the clustering degree improves, but the complexity of the clustering process increases
Solution Approach 1:
The patent implements a feedback mechanism where clustering parameters are calculated after each division step to evaluate the quality of clustering. This feedback information is used to guide subsequent division decisions, allowing the algorithm to automatically stop when further division no longer improves clustering quality, thus achieving high clustering degree without excessive complexity
Solution Approach 2:
The patent performs preliminary evaluation of clustering parameters before executing the full division process. By pre-calculating and comparing clustering metrics, the system determines the optimal number of divisions in advance, avoiding unnecessary computational complexity while ensuring high clustering degree is achieved
Data Source
AI summary
This application discloses an image clustering method performed by a computer device. This method includes: determining a first clustering parameter based on M image groups; for a target image group, dividing it into two image groups to obtain M+1 reference image groups, and determining a reference clustering parameter based on the M+1 reference image groups as a second clustering parameter for the target image group representing a clustering degree of images in the M+1 reference image groups; choosing, among respective second clustering parameters of the M image groups, a target second clustering parameter representing the highest clustering degree of images in the M image groups; and when the target second clustering parameter has a clustering degree not lower than the first clustering parameter, dividing a target image group corresponding to the target second clustering parameter into two image groups to obtain M+1 image groups.