Image Discrimination Using Non-Common Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques struggle with accurately discriminating between similar product types due to the use of unsuitable information, leading to reduced accuracy in product type discrimination, especially when similar model images are registered.
Innovation Solution
An image processing device and method that identifies and utilizes non-common regions within model images, which have distinct feature amounts, to perform discrimination processing, allowing for more accurate type classification by focusing on specific portions of the input image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If all information of similar model images is used for discrimination processing, then the discrimination processing can be performed comprehensively, but the accuracy of discrimination processing deteriorates due to inclusion of unsuitable information
Solution Approach 1:
The patent extracts and isolates non-common regions from model images, which are regions containing feature amounts that are different from other objects. By taking out only these distinctive regions and using them for discrimination processing, the system achieves high accuracy while maintaining comprehensive coverage of discriminative features.
Solution Approach 2:
The patent applies local quality by focusing discrimination processing on specific local regions (non-common regions) rather than using all image information uniformly. Each non-common region is processed and weighted according to its discriminative value, allowing the system to prioritize regions with high differentiation power while excluding regions with low discriminative value.
2Adaptability or versatility
If similar model images are registered, then the system can handle diverse product types, but the discrimination accuracy deteriorates due to similarity between models
Solution Approach 1:
The patent extracts non-common regions from similar model images, which are regions where the models differ from each other. By focusing discrimination on these extracted distinctive regions rather than on the overall similar appearance, the system can accurately distinguish between diverse product types even when the models are visually similar.
Solution Approach 2:
The patent creates asymmetry in the discrimination process by selectively emphasizing non-common regions that break the symmetry between similar models. This allows the system to find discriminative cues in the asymmetric differences between models, enabling accurate classification despite overall similarity.
3Quantity of substance
If feature amounts from all regions are used for discrimination, then the processing can be performed using complete information, but the reliability of discrimination result deteriorates due to unsuitable information
Solution Approach 1:
The patent extracts and isolates non-common regions containing reliable feature amounts from the complete model images. By taking out only these reliable regions and using them for discrimination, the system maintains high reliability of results while still processing comprehensive information about the objects.
Solution Approach 2:
The patent segments the model images into common regions and non-common regions, processing each segment differently. The non-common regions are selected for discrimination based on their reliability as discriminative features, while common regions are excluded. This segmentation allows the system to maintain reliability by focusing on the most trustworthy information segments.
Data Source
AI summary
The invention provides an image processing device, an image processing method, and an image processing program capable of correctly discriminating a type of a test object, even when similar model images have been registered. The image processing device includes: a hardware that holds a feature amount obtained from model images of a plurality of reference objects belonging to mutually different types; a region determination module that determines a non-common region as a region indicating a feature amount different from those of other objects, within a model image of each object; and a discrimination module that discriminates which type an object included in an input image belongs to, by using a feature amount corresponding to a non-common region of the object out of feature amounts of objects.