Image Feature Matching Stabilizing Flat Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image matching techniques, such as those using Binary Robust Independent Elementary Features (BRIEF), face challenges in accurately determining similarity between feature points due to variations in light source intensity and noise, especially in flat regions of images, leading to poor accuracy in image recognition.
Innovation Solution
The proposed solution involves calculating local feature values by comparing luminance differences of pixel pairs with a specified range that includes zero, assigning fixed bit values (0 or 1) to pixel pairs within this range to stabilize bit values against variations, ensuring accurate similarity evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If binary code is used to represent local feature values based on pixel-to-pixel luminance differences, then high-speed evaluation of similarity between feature points is achieved, but accuracy of image recognition deteriorates in flat and monotonous regions due to uncertainty in local feature values
Solution Approach 1:
The patent applies local quality by differentiating the treatment of bit values based on their spatial context. Bit values corresponding to pixel pairs in flat or monotonous regions (where luminance differences are small) are treated differently from those in regions with significant luminance variations. This is achieved by identifying regions with small luminance differences and applying a stabilization mechanism (such as setting bit values to a predetermined value or using alternative computation) specifically for these regions, while maintaining the standard binary code approach for regions with sufficient contrast. This localized adaptation preserves high-speed evaluation for most regions while improving accuracy for problematic flat regions.
2Ease of manufacture
If bit values are calculated based on the sign of luminance differences of pixel pairs, then local feature values can be expressed in binary code form, but reliability of feature matching deteriorates when luminance differences are close to zero due to sensitivity to light source variations and noise
Solution Approach 1:
The patent applies parameter changes by modifying the computation method for bit values based on the magnitude of luminance differences. When luminance differences fall within a predetermined range (indicating flat or monotonous regions), the patent changes the parameter used for bit value determination. Instead of relying solely on the sign of small luminance differences (which is sensitive to noise and light variations), the system either sets bit values to a predetermined value or uses alternative computation methods. For pixel pairs with larger luminance differences outside this range, the standard binary code calculation based on the sign of luminance differences is maintained. This conditional parameter change improves reliability for problematic regions while preserving the simplicity and efficiency of binary code for regions where it works well.
Data Source
AI summary
A computation unit calculates luminance differences of individual pixel pairs in a feature area and calculates, based thereon, a local feature value formed from bit values respectively corresponding to the pixel pairs. Specifically, the computation unit calculates a specific luminance difference for a specific pixel pair corresponding to a specific bit value and then compares the result with a specified range including a zero point of luminance difference. Then a first value is assigned to the specific bit value when the specific luminance difference is greater than the upper bound of the specified range. A second value is assigned to the same when the specific luminance difference is smaller than the lower bound of the specified range. A predetermined one of the first and second values is assigned to the same when the specific luminance difference falls in the specified range.


