Hierarchical Image Classification Using Adaptive Resonance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image search technologies are inefficient and inaccurate due to reliance on metadata and limited processing speed, resulting in poor search results and limited applicability to large image collections.
Innovation Solution
A method and system for image classification using feature descriptors, which generates classifiers to create a hierarchical classification structure within an image database, enabling efficient search by determining simple, complex, and hypercomplex feature descriptors and performing adaptive resonance classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current image search technologies use metadata or simple image features for searching, then the search process is simple and fast, but the search accuracy and relevance are poor
Solution Approach 1:
The patent segments image features into multiple hierarchical levels: simple features (edges, corners), complex features (textures, shapes), and hypercomplex features (object-level descriptors). This segmentation allows the system to process images at different levels of detail, improving search accuracy without overwhelming the system with all features at once.
Solution Approach 2:
The patent introduces a hierarchical dimension to the classification structure, organizing features from simple to hypercomplex across multiple levels. This dimensional organization transforms the flat, single-level classification into a multi-layered hierarchy, enabling more accurate image matching while managing complexity through structured organization.
2Measurement precision
If hierarchical classification structure is implemented to improve search accuracy, then image search relevance improves, but processing speed decreases
Solution Approach 1:
The patent performs preliminary classification of images into hierarchical categories before executing detailed image searches. By pre-organizing images into broad categories (e.g., natural scenes, man-made objects) and subcategories, the system quickly eliminates irrelevant images early in the search process, reducing the number of images requiring detailed feature comparison and thus improving overall processing speed.
Solution Approach 2:
The patent applies partial classification action by focusing computational resources on comparing features only within relevant hierarchical categories rather than all images in the database. This partial action approach processes only the necessary subset of images at each hierarchical level, maintaining high search relevance while optimizing processing speed by avoiding unnecessary comparisons.
3Measurement precision
If feature descriptors are grouped into complex andhypercomplex features, then image classification accuracy improves, but computational complexity increases
Solution Approach 1:
The patent segments feature processing into distinct computational stages: extracting simple features, grouping them into complex features, and combining complex features intohypercomplex features. This segmentation of the computational process allows each stage to be optimized independently, managing overall complexity while achieving high classification accuracy through progressive feature integration.
Data Source
AI summary
An electronic image classification and search system and method are provided. Images are processed to determine a plurality of simple feature descriptors based upon characteristics of the image itself. The simple feature descriptors are grouped into complex features based upon the orientation of the simple feature descriptors. End-stopped complex feature descriptors and complex feature descriptors at multiple orientations are grouped into hypercomplex feature descriptors. Hypercomplex resonant feature descriptor clusters are generated by linking pairs of hypercomplex feature descriptors. Feature hierarchy classification can then be performed by adaptive resonance on feature descriptors and classifier metadata associated with the image can then be generated to facilitate indexing and searching of the image within a hierarchical image database.