Defective Pixel Detection Using Statistical Distance Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting defective pixels in image arrays are inefficient, often misclassifying pixels as 'noisy' or failing to identify pixels that are not at intensity extremes, and may require costly computational resources or physical calibration.
Innovation Solution
A method for on-the-fly detection of defective pixels using statistical distribution analysis, where each pixel is analyzed within a selected cell of pixels, determining an allowed statistical distance, and identifying pixels exceeding this distance as defective.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If filtering is applied across the image to handle salt and pepper noise, then noise is reduced, but functional pixels are also adjusted resulting in unwanted smoothing of textures or removal of features
Solution Approach 1:
The patent performs preliminary identification and marking of defective pixels using statistical analysis before applying any filtering operations. By pre-marking only the actual defective pixels based on their statistical deviation from neighboring pixels, the system ensures that subsequent filtering operations are applied selectively only to marked pixels rather than across the entire image, thus preserving textures and features in functional pixels.
2Measurement precision
If peer group filtering methods are used for defective pixel detection, then detection accuracy is improved, but computational cost increases requiring multiple stages of calculation and histograms
Solution Approach 1:
The patent changes the approach from complex peer group filtering using histograms and multiple calculation stages to a simpler statistical parameter-based method. It uses basic statistical parameters (mean, standard deviation) calculated from a small local neighborhood of pixels to detect defective pixels. This parameter change dramatically reduces computational complexity while maintaining detection accuracy, as it avoids the need for complex histogram calculations and iterative peer group comparisons.
3Measurement precision
If preliminary calibration methods are used to mark defective pixels, then detection capability is improved, but physical array modification is required which is not always possible
Solution Approach 1:
The patent creates a digital copy or map of the defective pixel locations through statistical analysis of the image data, rather than requiring physical modification of the sensor array. The system generates a defective pixel mask or lookup table that stores the coordinates of identified defective pixels, allowing software-based correction without any physical changes to the hardware. This copying approach makes the solution universally applicable to any imaging system without requiring physical calibration capabilities.
4Reliability
If pixels are permanently marked as defective, then detection reliability is improved, but momentary defective pixels are incorrectly marked leading to counter-productive results
Solution Approach 1:
The patent implements periodic re-evaluation of pixel status by continuously monitoring statistical parameters and updating the defective pixel mask dynamically. Instead of permanent marking, the system periodically reassesses pixels based on current image data, allowing momentary defective pixels to be identified and corrected temporarily without permanent damage to the pixel's functional status. This periodic action enables the system to adapt to changing conditions and avoid false positives from transient issues.
Data Source
AI summary
Described herein are systems and methods for on-the-fly detection of a defective pixel in an image that may include: receiving the image; for each analyzed pixel in the image, selecting an analyzed cell, wherein the analyzed cell includes a plurality of pixels; performing statistical distribution analysis for the plurality of pixels in the analyzed cell; determining an allowed statistical distance for the analyzed pixel relative to the analyzed cell; and determining that the analyzed pixel is a defective pixel if a computed statistical distance of the analyzed pixel exceeds the allowed statistical distance.