Image Tagging via Similarity-Based Tag Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital content tagging systems rely heavily on user-provided tags, which are error-prone and time-consuming, and automatic tagging methods like object recognition are processor-intensive and often focus on unimportant image features.
Innovation Solution
The system compares an image to a set of pre-tagged images using photogrammetric techniques to extract and combine prominent tags, ordering them by importance for efficient tagging, even when the image is initially untagged or has existing tags.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user-provided tags are used for digital content tagging, then tagging accuracy can be maintained, but the process becomes extremely time-consuming and error-prone
Solution Approach 1:
The patent copies tags from similar pre-tagged images to the target image. Instead of requiring users to manually tag each image, the system identifies visually similar images from the database and transfers their tags to the new image, dramatically reducing tagging time while maintaining accuracy through the use of verified tags from similar images.
Solution Approach 2:
The system performs preliminary tagging by automatically identifying similar images and extracting their tags before user intervention is needed. This preliminary action prepares the target image with relevant tags automatically, eliminating the need for time-consuming manual tagging while ensuring accuracy through pre-verified tags from similar images.
2Productivity
If object recognition is used for automatic tagging, then tagging time is reduced, but the process becomes processor and memory intensive
Solution Approach 1:
The patent extracts only the necessary tags from similar pre-tagged images rather than performing complete object recognition on the target image. By taking out and reusing existing tags from visually similar images, the system achieves automatic tagging with minimal processor and memory consumption, avoiding the intensive computational requirements of full object recognition.
3Loss of information
If object recognition is used to identify all recognizable items, then comprehensive tagging is achieved, but many unimportant features are tagged reducing relevance
Solution Approach 1:
The patent applies local quality by selectively transferring tags based on the visual similarity between images. Instead of copying all tags uniformly, the system evaluates the degree of similarity and transfers only the relevant tags that are appropriate for the target image, ensuring high tag relevance while maintaining comprehensive coverage of important features.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach automates the tagging process, improving accuracy and efficiency by leveraging pre-existing tags from similar images, enhancing search functionality and reducing user input errors, while focusing on relevant features.
Implementation Method 1
The comparison can be based upon a photogrammetric technique, such as that used for photo tourism or implemented within MS PHOTOSYNTH
Data Source
AI summary
An image can be compared with a set of images, each including pre-existing tags. A similar image set can be determined from results of the comparing. Pre-existing tags can be extracted from the similar image set. Prominent tags can be determined from the extracted pre-existing tags. At least one of the determined prominent tags can be added to a tag set associated with the image.