Infrared Pixel Detection via Laplace Score Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Infrared image capture devices face challenges in detecting and correcting parasitic pixels, especially in shutterless devices, where image degradation occurs due to miscalibrated pixels, and existing methods are inefficient or require costly mechanical shutters.
Innovation Solution
A method for detecting parasitic pixels by calculating scores based on distances from neighboring pixels within a defined window, using a threshold calculation based on the Laplace distribution, and updating the operability map to correct miscalibrated pixels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a mechanical shutter is used to identify parasitic pixels, then detection accuracy is improved, but device complexity and cost increase
Solution Approach 1:
The patent extracts the shutter function from the physical mechanical component and implements it through software-based frame capture and processing. The system captures multiple frames with the sensor array exposed, effectively removing the need for a mechanical shutter while maintaining the ability to identify parasitic pixels through frame comparison algorithms.
Solution Approach 2:
The patent replaces the mechanical shutter system with an electronic/software-based solution. Instead of physically blocking the sensor, the system uses electronic frame capture and digital image processing to achieve parasitic pixel identification, substituting mechanical movement with electronic control and computational analysis.
2Measurement precision
If a mechanical shutter is used for calibration, then parasitic pixel detection is improved, but productivity decreases due to time loss
Solution Approach 1:
The patent enables continuous image capture without interruption by removing the mechanical shutter's closing phase. The sensor array remains continuously exposed to the scene, capturing multiple frames in sequence without blocking, thereby maintaining uninterrupted useful action while still enabling parasitic pixel detection through frame comparison.
Solution Approach 2:
The system performs parasitic pixel detection by analyzing relationships between multiple captured frames, effectively using preliminary frame data to identify and correct parasitic pixels in real-time during normal operation, rather than requiring a separate calibration phase with shutter closing.
3Ease of operation
If simple neighbor-based replacement is used for bad pixels, then ease of operation is improved, but image quality deteriorates in textured areas
Solution Approach 1:
The patent changes the approach from simple spatial neighbor replacement to a more sophisticated method that considers multiple parameters including temporal variations across frames, spatial relationships, and statistical properties. The system uses frame differencing and statistical analysis to distinguish parasitic pixels from actual scene features, maintaining image quality while correcting defects.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The invention relates to a method for detecting bad pixels from a pixel array of a device, for capturing an image, that is sensitive to infrared radiation. The method includes: receiving an input image captured by the pixel system, and calculating a score for a plurality of target pixels including at least some of the pixels from the input image. The score for each target pixel is generated on the basis of k pixels of the input image that are selected in a window of H by H pixels around the target pixel. H is an odd integer greater than or equal to 3, and k is an integer between 2 and 5. Each pixel, from the set formed of the k pixels and the target pixel, share at least one border or corner with another pixel from said set, and the values of the k pixels are at respective distances from the value of the target pixel, the k pixels being selected on the basis of the k distances. The method also includes detecting that at least one of the target pixels is a bad pixel on the basis of the calculated scores.