Defect Pixel Correction in Image Sensors Using Median Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image sensor technologies face challenges in effectively correcting defective pixels, particularly in color image sensors, as conventional methods either require additional memory, remove useful detail, or fail to distinguish between defect pixels and scene deviations, leading to issues like false color information and inability to handle intermittent pixel faults.
Innovation Solution
A method and apparatus for defect pixel correction in image sensors that uses a 7-pixel kernel to generate extrapolated values from neighboring pixels, applying a scaling factor to correct defective pixel values, and calculating a median value to determine the corrected pixel value, suitable for both color and black and white image sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional filtering methods (spatial median filter, Kalman filter) are used to remove isolated pixel faults, then defective pixels are corrected, but useful image detail is also removed
Solution Approach 1:
The patent applies different processing strategies to different pixel types. Defective pixels are identified and corrected using median filtering, while non-defective pixels (including bright stars) are preserved. The correction is localized only to pixels that fail the defect test, maintaining local quality differences between defective and non-defective pixels.
Solution Approach 2:
The patent performs preliminary identification of defective pixels before applying correction. A defect test is conducted on each pixel to determine if it is defective before applying the median filter. This preliminary action prevents unnecessary correction of non-defective pixels, preserving image detail and avoiding loss of information.
2Reliability
If memory is added to store defective pixel positions for correction, then isolated pixel values can be replaced, but silicon area and manufacturing cost increase
Solution Approach 1:
The patent makes the image sensor self-sufficient by performing defect identification and correction entirely within the sensor using on-chip logic circuits. No external memory or separate correction device is needed. The sensor identifies defective pixels through built-in test logic and corrects them using median filtering computed by on-chip arithmetic logic units, eliminating the need for additional memory components.
3Loss of information
If conventional demosaicing is used, then color information is reconstructed, but defect pixel information spreads to create false color in neighboring pixels
Solution Approach 1:
The patent extracts and removes defective pixel values from the demosaicing process before color reconstruction occurs. By identifying and correcting defective pixels prior to demosaicing, the harmful defect information is separated and eliminated, preventing it from spreading to neighboring pixels during the color interpolation process.
Solution Approach 2:
The patent performs defect correction as a preliminary step before demosaicing. The defect test and median filtering are applied to raw pixel values before the color reconstruction process begins. This preliminary action ensures that only corrected, non-defective values participate in demosaicing, preventing false color generation.
4Reliability
If a 7-pixel kernel with extrapolated values and median calculation is used, then defective pixels are corrected without additional memory, but computational complexity increases
Solution Approach 1:
The patent segments the correction process into distinct functional modules: defect testing, kernel extraction, extrapolated value calculation, and median computation. Each module is implemented by dedicated logic circuits within the image sensor, distributing the computational complexity across separate functional blocks rather than requiring a single complex processing unit.
Data Source
AI summary
A method and apparatus for defect pixel correction in an image sensor. The method may include calculating a median of extrapolated values of right neighboring and left neighboring pixel values of a potentially defective pixel.