Random Noise Evaluation for High-Resolution Image Sensors

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image sensor noise evaluation methods require large memory capacity, especially as the number of pixels increases, making them inefficient for high-resolution image sensors.

Innovation Solution

A method and device for evaluating random noise using a calculation unit that calculates differences and square sums between current and previous image data, allowing for the determination of random noise values with minimal memory storage by using a series of image data frames, thereby optimizing memory usage.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a general RN evaluation algorithm is used that calculates standard deviation and RMS of all pixels, then accurate random noise evaluation is achieved, but memory size increases by geometric progression as the number of pixels increases

Engineering Contradiction:

Improverandom noise evaluation accuracyVSAvoidmemory capacity

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the necessary computational elements for noise evaluation (difference values and their squares) from the complete image data set. Instead of storing and processing all pixel values, the method extracts and accumulates only the difference squares needed for variance calculation, thereby reducing memory requirements while maintaining evaluation accuracy

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the noise evaluation process into incremental updates. Rather than computing all differences and storing complete image frames, the method segments the calculation into sequential frame-by-frame updates of the difference square sum, allowing memory-efficient processing where only cumulative statistics are retained

Inventive Principle:

Principle #1Segmentation

2Manufacturing precision

If image sensor resolution is increased by scaling down photodiodes, then higher resolution is achieved, but noise characteristic deteriorates due to lower optical signal level

Engineering Contradiction:

Improveimage sensor resolutionVSAvoidnoise characteristic

Core Design Contradiction:

Manufacturing precisionVSReliability

Solution Approach 1:

The patent replaces physical/optical improvements (increasing photodiode size to improve signal level) with a computational approach. Instead of modifying the optical system to increase signal, the method uses statistical processing of multiple frames to extract noise characteristics, substituting mechanical/optical solutions with information-processing solutions

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS7834918B2Method and device for evaluating random noise in image sensor

Publication Date: 2010.11.16 SAMSUNG ELECTRONICS CO LTD

AI summary

A random noise evaluation method is comprised of: capturing and obtaining current image data; calculating a difference between the current image data and an average of previous image data; calculating a current difference square sum using a sum of the difference and a previous difference square sum; and calculating a random noise value using the current difference square sum. Since a noise evaluation algorithm obtains random noises without storing image data for all pixels of a*b*n, it is able to evaluate random noise evaluation for a high-resolution image sensor even with a relatively small size of memory.