Sketch Query Image Search System Using Visual Attribute Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Text-based search systems are inadequate for efficiently managing and searching vast multimedia data, particularly images, as they struggle to represent visual information effectively.
Innovation Solution
An image search system that utilizes a query engine to reconfigure search queries based on descriptions and attributes of sketch queries, performing logical operations on sketch queries of the same or different colors, and includes a sketch input unit for intuitive input tools to facilitate user interaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If text-based search is used for image retrieval, then the system is simple to implement, but it cannot effectively represent visual information
Solution Approach 1:
The system segments the image search problem into two parts: visual feature extraction (using color histograms, texture features, shape descriptors) and text-based query processing. This allows the system to maintain text-based search simplicity while incorporating visual information representation through separate feature analysis modules.
Solution Approach 2:
The patent introduces visual feature vectors as an intermediary between the text-based query system and the image database. These feature vectors (comprising color, texture, and shape attributes) serve as a bridge that translates visual information into a format that can be processed by text-based search algorithms, thereby preserving visual information without complicating the overall search architecture.
2Measurement precision
If sketch queries with multiple attributes are used to improve search accuracy, then the search precision increases, but the system complexity increases
Solution Approach 1:
The system segments complex sketch queries into distinct visual attribute components (color, texture, shape, size). Each attribute is processed independently through dedicated feature extraction modules, and the results are combined to form a comprehensive search query. This segmentation maintains search accuracy while managing system complexity through modular processing.
Solution Approach 2:
The patent transforms visual attributes from qualitative sketch descriptions into quantitative parameter representations (e.g., color histograms with specific bins, texture co-occurrence matrices, shape factor values). This parameter transformation enables precise measurement and comparison while providing a standardized framework that simplifies the processing of complex multi-attribute queries.
3Productivity
If the search space is reduced to improve search speed, then the productivity increases, but the search completeness may be compromised
Solution Approach 1:
The system performs preliminary indexing of visual features (color histograms, texture descriptors, shape parameters) for all images in the database before actual search operations. This pre-processing creates an organized structure of visual attributes that enables rapid filtering and comparison during search, significantly improving speed while maintaining completeness through comprehensive feature coverage.
Solution Approach 2:
The patent transforms the high-dimensional visual feature space into a reduced-dimensional representation using techniques such as principal component analysis or other dimensionality reduction methods. This creates a compressed feature space that retains the essential visual information needed for accurate matching while reducing the computational complexity and search space, thereby improving search speed without sacrificing completeness.
Data Source
AI summary
Provided are image search system and methods. The image search system may include a query engine for reconfiguring a search query based on a description and an attribute of a sketch query that is input to a query image; a search engine for extracting image data that matches the search query from a database; and a display apparatus for displaying the extracted image data.