Pixel Value Correction Circuitry for Row Correlated Noise Mitigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Imaging devices face challenges in mitigating row correlated noise, such as temporal row noise, which results in unsightly flickering in captured images due to noise offsets across rows, and quantization errors during noise correction.
Innovation Solution
The implementation of pixel value correction circuitry that includes dark pixels to estimate noise components and generate correction values, with extended precision and dithering techniques to mitigate noise and quantization errors, ensuring accurate noise removal without introducing visible artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If standard precision correction values are used to remove row noise, then noise mitigation is achieved, but quantization errors introduce visible artifacts
Solution Approach 1:
The patent introduces an additional precision dimension by implementing extended precision arithmetic (e.g., 20-bit precision) beyond the standard 13-bit pixel data precision. This dimensional expansion in the precision domain allows correction values to be calculated and applied with sufficient granularity to avoid quantization errors, while still being representable in standard formats for final output.
Solution Approach 2:
The patent changes the precision parameter of correction values from standard precision to extended precision during intermediate calculations. By adjusting this parameter, the system maintains high accuracy in noise correction computations, then restores standard precision for final pixel values, thereby eliminating visible quantization artifacts while preserving noise mitigation effectiveness.
2Manufacturing precision
If extended precision correction values are used, then quantization errors are mitigated, but device complexity increases
Solution Approach 1:
The patent segments the processing pipeline into distinct functional blocks: a precision extension block that converts standard precision values to extended precision, a correction computation block that performs noise subtraction, and a precision restoration block that converts results back to standard precision. This segmentation allows extended precision to be used only where necessary, minimizing overall system complexity.
Solution Approach 2:
The patent introduces an intermediary extended precision representation as a temporary state during correction calculations. This intermediary format acts as a bridge between standard precision input data and standard precision output data, enabling high-precision computations without requiring the entire system to operate at extended precision, thus limiting complexity increase to specific processing stages.
3Manufacturing precision
If row noise correction is applied, then image quality improves, but processing time increases due to additional correction steps
Solution Approach 1:
The patent performs preliminary computation of correction values using extended precision arithmetic before applying them to pixel data. By pre-calculating correction values with sufficient precision, the system avoids the need for complex real-time computations during image processing, reducing overall processing time while maintaining high image quality.
Solution Approach 2:
The patent replaces complex, time-consuming high-precision computational operations with a streamlined sequence of simpler operations: standard precision multiplication, extended precision conversion, correction subtraction, and precision restoration. This substitution maintains accuracy while reducing computational complexity and processing time compared to performing all operations in extended precision.
Data Source
AI summary
An image sensor may have an array of pixels and readout circuitry. The array may include image pixels that generate signals in response to image light and reference pixels that generate signals in response to electrical noise. The readout circuitry may obtain first pixel values from the image pixels and may obtain second pixel values from the reference pixels. The readout circuitry may generate an extended precision pixel value based on the second pixel values that have an extended bit width relative to the each of the second pixel values. The readout circuitry may generate multiple dithered correction values by adding randomized sequences of least significant bits to the extended precision pixel value. The readout circuitry may mitigate visible quantization error and noise such as row-correlated and column-correlated noise in the final image by subtracting the dithered correction values from corresponding first pixel values.