Per-Pixel Correction Model for Fixed Pattern Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional black frame subtraction techniques for digital cameras are inefficient in reducing signal-dependent fixed pattern noise, require large memory storage, and cause camera startup delays due to the need for extensive black frame capture and processing, which is temperature-dependent and inaccurate at varying sensor temperatures.
Innovation Solution
A method that calibrates per-pixel correction models at different illumination levels and temperatures, using compressed signal-dependent and temperature-dependent correction models to reduce memory bandwidth and eliminate delays by calibrating during power down routines, thereby correcting fixed pattern noise and bad pixels effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If black frame subtraction is used to cancel fixed pattern noise, then signal-independent noise is reduced, but signal-dependent noise cannot be reduced and large memory storage is required
Solution Approach 1:
The patent segments the fixed pattern noise into signal-independent and signal-dependent components, and applies different correction methods to each. The signal-independent component is corrected using traditional black frame subtraction, while the signal-dependent component is corrected using illumination-level-specific correction values stored in memory, thereby reducing overall memory requirements while achieving comprehensive noise reduction.
Solution Approach 2:
The patent changes the parameter of correction data by storing multiple sets of correction values corresponding to different illumination levels rather than a single black frame. This allows the system to select the appropriate correction values based on current illumination conditions, reducing the need to store large black frame data for all possible lighting scenarios.
2Measurement precision
If black frame capture is performed at startup or between pictures, then fixed pattern noise is calibrated, but camera startup time and picture-to-picture time are increased
Solution Approach 1:
The patent performs preliminary calibration actions by capturing black frames at multiple illumination levels during factory calibration or initial setup, and pre-processing these into correction values that are stored in memory. This preliminary preparation eliminates the need for time-consuming black frame capture during camera startup or between pictures, as the correction values are already available for immediate use.
Solution Approach 2:
The patent implements a dynamic correction system that automatically selects the appropriate correction values based on real-time illumination level detection. Instead of performing static calibration at fixed intervals, the system dynamically adapts by choosing pre-computed correction values that match current lighting conditions, eliminating calibration delays while maintaining accuracy.
3Measurement precision
If black frame subtraction is used, then noise cancellation is achieved, but the black frame size is large requiring high-bandwidth memory
Solution Approach 1:
The patent extracts only the essential correction information from black frames by computing correction values that represent the fixed pattern noise characteristics at different illumination levels. Instead of storing and processing entire black frame images, the system extracts and stores compact correction values, dramatically reducing memory bandwidth requirements while preserving noise cancellation capability.
Solution Approach 2:
The patent inverts the traditional approach by not storing black frames directly for subtraction, but rather storing pre-computed correction values that represent the noise characteristics. This inversion transforms the process from direct black frame subtraction requiring high bandwidth to efficient correction value application requiring minimal bandwidth.
Data Source
AI summary
A method of pixel correction is disclosed. The method generally includes the steps of (A) calibrating a per-pixel correction model of a sensor at a plurality of different illumination levels, (B) generating a plurality of pixel values from the sensor in response to an optical signal and (C) generating a plurality of corrected values by applying the per-pixel correction model to the pixel values.