Hierarchical Image Clustering for Portable Device Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The hierarchical clustering algorithm for managing and displaying large volumes of image data is computationally intensive, leading to performance deterioration in portable devices, making it difficult to efficiently search and filter image data based on location and time.
Innovation Solution
An image display apparatus with hierarchical clustering that includes image information storage, cluster managing, display layer designation, conditional filtering, and map storage, allowing for intuitive data management and filtering by designating display conditions such as location and time ranges, and displaying image locations on a map.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If hierarchical clustering algorithm is applied to manage image data, then image data can be arranged hierarchically by location, but calculation quantity increases rapidly with square order of image data
Solution Approach 1:
The patent divides the image data set into multiple clusters based on location proximity, organizing data into a tree structure with parent-child relationships. This segmentation reduces the computational burden by processing smaller subsets of data rather than performing exhaustive one-to-one distance calculations across all image pairs.
Solution Approach 2:
The system pre-calculates and stores distance information between image data points during the clustering process. By performing this calculation in advance and caching the results, the system avoids repeating expensive distance computations during subsequent search and filtering operations, significantly improving retrieval speed.
2Measurement precision
If one-to-one distance calculation is performed for every image data, then accurate clustering is achieved, but calculation ability is exceeded in portable equipment
Solution Approach 1:
The patent implements a distance threshold mechanism where only image pairs within a certain distance range are subjected to detailed one-to-one distance calculation. Image pairs exceeding the threshold are quickly filtered out using bounding box or approximate location comparisons, reducing the number of expensive precise calculations required while maintaining acceptable clustering accuracy.
3Quantity of substance
If all image data is displayed without filtering, then complete data is shown, but user search efficiency is reduced
Solution Approach 1:
The patent introduces an intermediary filtering layer between the user and the full image database. This filter uses stored location and time metadata to quickly identify and retrieve only the subset of images matching user-specified criteria, acting as a mediator that reduces the data volume presented to the user while maintaining access to the complete underlying dataset.
Data Source
AI summary
An image display apparatus includes: an image information storage that stores image information including locations where images are captured as information corresponding to captured image data; a cluster managing section that manages the image information with hierarchical clusters including a plurality of clusters configured in a plurality of layers; a display layer designation section that designates a layer to be displayed as a display layer among from an uppermost layer through a lowermost layer in the plurality of layers; a display condition designation section that designates a display condition of the image data; a conditional filtering section that selects such image data in the display layer that the image information thereof exists within a predetermined range based on the display condition; and a drawing section that draws in the display layer, as an image location display, the image captured locations of the image data selected based on the image information.