Parallel Arithmetic Mean Modules for Imager Noise Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional CMOS image sensors face challenges in efficiently generating and applying noise correction values due to significant calculation time and memory resource requirements, particularly in reducing fixed pattern and row noise, which leads to delays in readout and high memory capacity needs.

Innovation Solution

The method involves calculating pixel offset values and average pixel output values during calibration by determining differences between individual pixel outputs and average values, averaging these values across columns and rows, and using parallel arithmetic mean modules to facilitate efficient noise reduction, thereby reducing memory requirements and calculation time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional calibration processes are used to remove fixed pattern noise by measuring output based on known optical input and computing pixel offsets, then noise correction accuracy is improved, but calculation time and memory resource requirements increase significantly

Engineering Contradiction:

Improvenoise correction accuracyVSAvoidcalculation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the pixel array into multiple blocks and processes each block independently with dedicated arithmetic mean modules. This parallel processing approach divides the overall calibration task into smaller concurrent sub-tasks, reducing total calculation time while maintaining correction accuracy for each block

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary calculation of average pixel output values during the calibration phase and stores these pre-computed values in memory. During normal operation, only simple subtraction operations are needed to apply correction, significantly reducing real-time calculation time while preserving noise correction accuracy

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If conventional calibration processes compute and store pixel offset values for each pixel column, then fixed pattern noise correction is achieved, but memory capacity requirements increase

Engineering Contradiction:

Improvefixed pattern noise correctionVSAvoidmemory capacity

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential correction information (average pixel output values per block) and stores these compact representations in memory instead of storing individual pixel offset values. This extraction approach maintains the ability to correct fixed pattern noise while dramatically reducing memory capacity requirements

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

By dividing the pixel array into blocks and computing a single average value per block rather than per-pixel offsets, the patent segments the memory storage requirement into fewer, more compact data structures, reducing overall memory capacity needs while preserving correction effectiveness

Inventive Principle:

Principle #1Segmentation

3Measurement precision

If noise reduction processes are applied sequentially to serial data stream in conventional sensors, then fixed pattern noise and row noise are corrected, but readout delay increases

Engineering Contradiction:

Improvenoise reduction effectivenessVSAvoidreadout speed

Core Design Contradiction:

Measurement precisionVSSpeed

Solution Approach 1:

The patent implements parallel arithmetic mean modules that process multiple pixel blocks simultaneously rather than sequentially. This parallelization maintains noise reduction effectiveness for both fixed pattern and row noise while significantly increasing readout speed by eliminating sequential processing bottlenecks

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs the computationally intensive averaging calculations during the calibration phase and stores results for rapid application during normal readout. This preliminary computation separates the heavy processing from the time-critical readout path, maintaining noise reduction effectiveness while improving readout speed

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS7830423B2Method and apparatus providing imager noise reduction using parallel input arithmetic mean modules

Publication Date: 2010.11.09 APTINA IMAGING CORP

AI summary

Noise correction values are generated through the calculation of pixel offset values and average pixel output values during a calibration procedure. For each read calibration row, an average pixel output value is determined for a plurality of pixel output values. An offset value for each of the plurality of pixel outputs is generated by finding the difference between each of the plurality of pixel outputs and the calculated average pixel output value. An offset value for an entire row of pixel outputs may also be calculated by finding the average pixel output value for a plurality of optically black pixels in the imager row. The average pixel output value for the optically black pixel values may be refined by discarding pixel values outside a threshold window and substituting the average pixel output value and calculate a new average pixel value.