Content-Based Image Retrieval Using Visual Feature Descriptors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for retrieving images from large collections, such as user tags and metadata, are often ineffective due to the time-consuming manual process of tagging and the limitations of tag-based search techniques.
Innovation Solution
A content-based image searching system that uses feature descriptors and machine learning to generate classifiers, allowing users to interactively select positive and negative examples to rank images based on visual attributes, enabling effective retrieval of desired images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If tag-based search techniques are used to retrieve images, then images can be organized and searched using keywords, but the process becomes ineffective for large collections due to the time-consuming manual tagging process
Solution Approach 1:
The patent replaces the manual mechanical process of tagging images with an automated content-based visual search system. Instead of manually assigning keywords to images, the system uses computer vision algorithms to automatically analyze image content, extract visual features, and retrieve relevant images based on query images, thereby eliminating the time-consuming manual tagging process while maintaining effective image retrieval
Solution Approach 2:
The system enables images to serve themselves by automatically generating visual descriptors and organizing image collections based on their inherent visual content rather than requiring external manual annotation. The visual search system autonomously processes images, creates feature representations, and performs retrieval operations without human intervention in the tagging process
2Quantity of substance
If the collection of images is increased to accumulate more digital images, then more images are available for selection, but finding a desired image becomes more difficult
Solution Approach 1:
The patent transforms the search approach by changing from keyword-based parameters to visual feature parameters. Instead of searching through text metadata, the system converts images into visual feature spaces using computer vision techniques, allowing efficient retrieval even as the number of images increases. This parameter transformation enables the system to handle large collections by operating in a dimensional space defined by visual characteristics rather than textual descriptors
Solution Approach 2:
The system introduces a new dimensional approach to image retrieval by using multi-dimensional visual feature spaces. Rather than linear text-based search, images are represented in high-dimensional spaces defined by color histograms, texture features, shape descriptors, and other visual attributes, enabling efficient navigation and retrieval through geometric operations in this expanded dimensional domain
3Ease of operation
If user tags are manually associated with images to improve searchability, then images can be organized by keywords, but the manual process is too time-consuming for practical use
Solution Approach 1:
The patent replaces manual tagging operations with automated visual analysis. The system uses computer vision algorithms to automatically extract meaningful descriptors from image content, eliminating the need for manual keyword assignment while providing superior organization based on actual visual characteristics rather than subjective user labeling
Solution Approach 2:
Instead of manually creating text tags, the system automatically generates visual feature representations that copy and encode the essential visual characteristics of images. These visual descriptors serve as automated equivalents to manual tags, capturing image content in a structured format that enables efficient search and organization without human intervention
Data Source
AI summary
An image retrieval program (IRP) may be used to query a collection of digital images. The IRP may include a mining module to use local and global feature descriptors to automatically rank the digital images in the collection with respect to similarity to a user-selected positive example. Each local feature descriptor may represent a portion of an image based on a division of that image into multiple portions. Each global feature descriptor may represent an image as a whole. A user interface module of the IRP may receive input that identifies an image as the positive example. The user interface module may also present images from the collection in a user interface in a ranked order with respect to similarity to the positive example, based on results of the mining module. Query concepts may be saved and reused. Other embodiments are described and claimed.