Image Orientation Determination Using Rotated Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image orientation determination methods are prone to errors when applied to images where a semantic rule cannot be easily set up, leading to incorrect orientation corrections, especially when feature changes are significant or when images have compositions inappropriate for orientation determination.
Innovation Solution
An image orientation determination device that extracts features from input images and compares them to positive and negative training images, using similarity calculations to decide on orientation determination based on threshold values, and corrects orientation based on orientation-specific features when appropriate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing image orientation determination methods are used, then orientation can be determined for images with clear semantic features, but determination accuracy deteriorates for images where semantic rules cannot be easily set up
Solution Approach 1:
The patent creates virtual rotated copies of the input image (90度, 180度, 270度 rotations) and compares them against a database of training images with known orientations. Instead of relying on semantic rules that may not apply to all image types, the system copies the image in different orientations and finds the best match through similarity calculation, making the method universally applicable while maintaining accuracy
Solution Approach 2:
The system calculates similarity scores between the input image (and its rotated versions) and training images, then uses this feedback to determine the most likely orientation. The similarity calculation provides quantitative feedback that guides the orientation determination, allowing the system to adapt to different image types based on empirical matching rather than predefined rules
2Measurement precision
If training image comparison method is used, then orientation determination can be performed, but determination accuracy deteriorates when feature changes are significant or images have inappropriate compositions
Solution Approach 1:
The patent dynamically rotates the input image by 90, 180, and 270 degrees to generate multiple versions for comparison. This dynamic transformation allows the system to adapt to images with significant feature changes or inappropriate compositions by finding the rotation that produces the best match with training images, thereby maintaining precision and reliability across diverse image types
Solution Approach 2:
The system pre-rotates the input image into multiple candidate orientations before performing similarity comparison with training images. This preliminary action of creating rotated versions ensures that regardless of the original image's composition or feature stability, the system has already prepared all possible orientation variants for accurate matching
Data Source
AI summary
When positive image similarity (degree of training image similarity between input image features and those of positive training image) is higher than a predetermined first threshold, image orientation determination decision section determines input image orientation. When negative image similarity (degree of training image similarity between input image features and those of a negative training image) is not lower than a predetermined second threshold value, the image orientation determination decision section does not determine input image orientation. When the image orientation determination decision section determines the orientation of the input image, image orientation determination section calculates orientation similarity reflecting similarity between input image features and those stored in orientation-specific features storage section. If the calculated orientation similarity satisfies a predetermined condition, the image orientation determination section determines input image orientation according to positive training image orientation related to the image features stored in the orientation-specific features storage section.