Image Segment Annotation via Graph Isomorphism
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies fail to efficiently refine the referenced area within an image using tagged information and do not normalize data effectively for query by example image searches, lacking integration of common word graph elements important for image-based queries.
Innovation Solution
A method involving a coordinate grid projection onto images to convert lines into a tree structure, analyzing tagged elements, and using graph isomorphism to identify and refine the image segments, allowing for better categorization and query refinement by matching image elements with textual information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If annotation schemes are applied to images and videos, then indexing and searching efficiency is improved, but the connection between annotations and specific image regions is lost
Solution Approach 1:
The patent segments the image into multiple regions and creates separate annotation entries for each region. Each annotation is linked to its specific image region through region identifiers, allowing efficient indexing while preserving the connection between annotations and their corresponding image areas.
Solution Approach 2:
The patent introduces an intermediary data structure that bridges annotations and image regions. This intermediary layer contains region identifiers that link annotations to specific image regions, enabling both efficient searching and accurate region association without direct coupling.
2Speed
If image processing is performed without integrating textual information, then processing speed is maintained, but the accuracy of query by example searches deteriorates
Solution Approach 1:
The patent performs preliminary processing of textual information by creating word graphs and extracting common elements before the actual image search. This preprocessing step organizes textual data into structured formats that can be quickly compared with image regions during search operations, maintaining speed while improving accuracy.
Solution Approach 2:
The patent transforms textual information into a different parameter format (word graphs with common elements) that can be efficiently compared with image features. This parameter transformation enables accurate text-image matching without significantly increasing processing time.
3Measurement precision
If common word graph elements are integrated with image elements, then the reference area identification accuracy is improved, but the system complexity increases
Solution Approach 1:
The patent extracts only the common elements from word graphs that are relevant to image regions, rather than processing entire textual documents. This extraction approach reduces the amount of data to be processed and decreases system complexity while maintaining accurate reference area identification.
Solution Approach 2:
The patent applies different processing strategies to different parts of the system: word graphs are processed to extract common elements, image regions are segmented and tagged, and their intersections are computed. Each component has specialized processing optimized for its specific function, improving accuracy without uniformly increasing overall complexity.
Data Source
AI summary
A system that compares the images submitted with a preprocessed database containing pictures, drawings, and patent drawings, among other media. The images are interrelated by comparing the content of the patent images, the narrative in the patents with the other visual media which may or may not be pre-tagged.