Binarized Image Generation with Local Thresholds for Defect Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image noise due to uneven lighting or sensor characteristics can cause over-detection of defects in binarized images, particularly in low-brightness regions, leading to erroneous judgments.
Innovation Solution
A binarized image generation method that calculates a reference value for each pixel based on surrounding brightness, sets a defect candidate region extraction threshold and an over-detection reduction threshold, and compares pixel brightness against both thresholds to generate a binarized image, suppressing noise-induced over-detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single threshold is used for binarization, then the defect detection is simple, but noise causes over-detection especially in low-brightness regions
Solution Approach 1:
The patent applies local quality by adapting the binarization threshold to each pixel's local characteristics. Instead of using a single global threshold, the system calculates a reference value from surrounding pixels and uses this to determine an appropriate threshold for each location, thereby suppressing noise-induced over-detection in low-brightness regions while maintaining simple operation through automated local adaptation
2Measurement precision
If the threshold is set low to detect all potential defects, then detection sensitivity increases, but false detection of noise increases
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the binarization threshold based on the local brightness characteristics of each pixel. The threshold is not fixed but changes according to the reference value calculated from surrounding pixels, allowing the system to maintain high sensitivity for actual defects while automatically raising the threshold in low-brightness regions to suppress false detection of noise
3Measurement precision
If local thresholding is applied to handle non-uniform brightness, then defect detection improves, but computational complexity increases
Solution Approach 1:
The patent applies local quality by calculating a reference value from surrounding pixels for each location, which captures the local brightness characteristics. This local adaptation approach improves defect detection accuracy in non-uniform lighting conditions while keeping the computational complexity manageable through straightforward neighborhood averaging rather than complex global optimization
4Object-generated harmful factors
If a high threshold is used to reduce noise, then false detection decreases, but actual defects in low-brightness regions may be missed
Solution Approach 1:
The patent applies parameter changes by making the threshold dynamic rather than static. The threshold changes locally based on the reference value calculated from surrounding pixels, allowing it to be higher in low-brightness regions to suppress noise while being appropriately lower in high-brightness regions to maintain complete defect detection, thus balancing both requirements
Data Source
AI summary
A binarized image generation method that generates a binarized image of a target object, wherein for each pixel in an image of the target object: a reference value calculation step calculates a reference value based on the brightness values of the pixels around the subject pixel; a defect candidate region extraction threshold calculation step calculates a defect candidate region extraction threshold by multiplying a certain constant by the reference value; an over-detection reduction threshold setting step sets a threshold; and a threshold comparison step determines whether the brightness value of the subject pixel is above both the defect candidate region extraction threshold and the over-detection reduction threshold, and then generates a binarized image of the target object, where the image is binarized into two regions: one region consisting of pixels whose brightness values are greater than or equal to both thresholds, and the other region.