IR Image Artifact Detection Using Histogram Gradient Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting IR illumination leakage artifacts in IR camera systems are unreliable in uncontrolled environments, particularly due to issues with variance analysis and edge detection, and are computationally demanding.
Innovation Solution
A method involving frame segmentation, histogram analysis, and regression line fitting is used to classify frame segments based on smooth gradients, distinguishing between artifacts and real objects without edge detection, suitable for uncontrolled environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If variance analysis of edge pixels is used to detect edge transition, then the method is simple to implement, but it produces false classifications because variance is an accumulative factor that loses information about transition
Solution Approach 1:
The patent replaces the statistical variance analysis method with a physics-based radiometric model that describes the actual physical process of light leakage through the gasket ring. Instead of using accumulative statistical measures that lose transition information, the model uses the relationship between pixel values and their spatial coordinates to directly characterize the smooth gradient of leakage artifacts, thereby resolving the contradiction between simplicity and reliability.
2Measurement precision
If ESF analysis is used to evaluate edge characteristics, then adequate results are provided in controlled environments, but it is not suited for uncontrolled environments where edges may not be identifiable or may have various shapes
Solution Approach 1:
The patent creates a universal detection model that works across all environments by abandoning the assumption of identifiable edges. The radiometric model relates pixel values directly to spatial coordinates and applies to any leakage artifact regardless of shape or location, making the method universally applicable whether the artifact is a small spot, a large region, or has an irregular shape, thus resolving the contradiction between precision and adaptability.
Solution Approach 2:
Instead of trying to identify edges and then analyze them (the traditional ESF approach), the patent inverts the logic by directly modeling the pixel value distribution across the entire artifact region without requiring edge detection. This inversion eliminates the problem of unidentifiable edges in uncontrolled environments while maintaining measurement precision.
3Reliability
If deep learning classification is used for artifact detection, then comprehensive pattern recognition is achieved, but the computational demands are too high and performance is unacceptable
Solution Approach 1:
The patent replaces the expensive deep learning computational resource with a simple, lightweight radiometric model that requires minimal computation. The model uses basic mathematical operations to relate pixel values to spatial coordinates, providing reliable artifact detection without the high computational demands of deep learning, thus resolving the contradiction between reliability and productivity.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The present disclosure relates to a computer implemented method for detecting artifacts in infrared (IR) illuminated IR images. Artifacts are separated from real objects by analyzing results of regression line fits to histograms of pixel values. Artifacts will have smooth transition gradients and therefore will show smaller deviations of the regression line fit to the histogram data.