Texture Representation via Dependency Relationships for Image Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing methods based on texture representation suffer from poor recognition accuracy due to limited texture information being reflected in the representation results.
Innovation Solution
An image processing method utilizing a neural network to obtain a dependency relationship between features of texture primitives based on direction information and a multi-scale feature map, which enhances the texture representation by incorporating both texture features and their dependencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional texture representation methods are used, then the processing is simple, but the image recognition accuracy is low
Solution Approach 1:
The patent segments the texture representation process into multiple components: extracting texture primitives, determining their types, establishing dependency relationships between them, and constructing a hierarchical structure. This segmentation allows the system to capture detailed texture information while maintaining organized processing, thereby improving recognition accuracy without overwhelming complexity
Solution Approach 2:
The patent introduces a new dimensional aspect to texture representation by incorporating dependency relationships between texture primitives. Instead of only representing texture features in traditional feature space, the method adds a relational dimension that captures how different texture primitives depend on each other, enabling more discriminative representation for improved recognition accuracy
2Loss of information
If multi-scale feature maps and direction information are incorporated, then texture information completeness improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-defining a limited set of texture primitive types and their characteristic features before processing the actual texture representation. This preliminary classification framework allows the system to efficiently organize multi-scale feature information and direction data without requiring complex real-time analysis, thus reducing computational complexity while maintaining information completeness
Solution Approach 2:
The patent applies local quality by processing different regions of the texture with appropriate levels of detail. Texture primitives are identified and classified based on local characteristics, and dependency relationships are established locally before being integrated into the global representation. This localized processing approach maintains comprehensive texture information while avoiding the need to process all features at maximum complexity uniformly across the entire image
Data Source
AI summary
An image processing method and apparatus, and a storage medium are provided, and relate to the image processing field. A dependency relationship between features of texture primitives of an image may be obtained based on direction information and a multi-scale feature map of the image, at least one group of texture features of the image may be obtained based on a feature map of the image on at least one scale, and a texture representation result of the image may be obtained based on the dependency relationship and the at least one group of texture features. Then, the image may be processed based on the texture representation result of the image. Because the texture representation result of the image can reflect more perfect texture information of the image, an image processing effect is better when image processing such as image recognition, image segmentation, or image synthesis is performed.