Image Search Query Expansion via Co-occurrence Keyword Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image retrieval systems face challenges in efficiently searching large personal image collections without manual annotations, as they rely on limited vocabularies and require users to match exact search terms, leading to either too few or too many results.
Innovation Solution
A method that indexes image collections using data processing systems to generate image descriptors, expands search queries by identifying co-occurrence keywords, and groups candidate images by descriptors for representative selection, enabling complex concept-based searches without user-provided annotations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotations are applied to images, then search accuracy is improved, but time consumption increases significantly
Solution Approach 1:
The system automatically generates annotations for images using computer vision algorithms and machine learning models, eliminating the need for manual annotation by users. The images self-annotate themselves through automated feature extraction and concept generation, resolving the contradiction between search accuracy and time consumption.
Solution Approach 2:
The patent replaces the mechanical process of manual annotation with automated computational processes including image processing, feature extraction, and algorithmic concept generation. This substitution of manual mechanical annotation with automated digital processing achieves both high search accuracy and efficiency.
2Ease of manufacture
If a limited vocabulary is used for image annotations, then annotation process is simplified, but search flexibility deteriorates
Solution Approach 1:
The system dynamically changes the parameter of vocabulary size and complexity based on the specific image and search context. Instead of using a fixed limited vocabulary, the system generates context-appropriate concepts and terms, allowing both simple and complex searches to be effective without compromising either annotation simplicity or search flexibility.
Solution Approach 2:
The annotation system transitions from static limited vocabularies to dynamic concept generation that adapts to each image's unique features. The vocabulary becomes flexible and context-dependent, allowing the system to maintain annotation simplicity while achieving search flexibility through adaptive concept creation.
3Measurement precision
If exact search term matching is required, then search precision is improved, but number of results decreases
Solution Approach 1:
The system introduces intermediary concepts and semantic relationships between search terms and image annotations. Instead of requiring exact matching, the intermediary semantic network allows for flexible matching while maintaining precision, enabling users to find relevant results even when search terms don't exactly match annotations through synonym recognition and concept mapping.
Solution Approach 2:
The search system dynamically adjusts the matching parameter from exact string matching to semantic similarity matching. This parameter change allows the system to maintain search precision through meaningful concept matching while increasing the number of relevant results by recognizing synonymous and related terms.
Data Source
AI summary
A method of identifying one or more particular images from an image collection, includes indexing the image collection to provide image descriptors for each image in the image collection such that each image is described by one or more of the image descriptors; receiving a query from a user specifying at least one keyword for an image search; and using the keyword(s) to search a second collection of tagged images to identify co-occurrence keywords. The method further includes using the identified co-occurrence keywords to provide an expanded list of keywords; using the expanded list of keywords to search the image descriptors to identify a set of candidate images satisfying the keywords; grouping the set of candidate images according to at least one of the image descriptors, and selecting one or more representative images from each grouping; and displaying the representative images to the user.