Image Processing via Spatial Frequency Decomposition for Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Increasing the throughput of image capture devices, such as document scanners, leads to reduced exposure time, resulting in increased image noise or the need for higher quality optics, which increases manufacturing costs.
Innovation Solution
An image processing method that utilizes three image planes from sensors with different spectral distributions, generating spatial frequency components and applying a color transform to combine them, thereby reducing noise and improving sensitivity without introducing noise into high-frequency components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the throughput of a document scanner is increased, then the productivity is improved, but the exposure time decreases resulting in increased image noise
Solution Approach 1:
The image processing is segmented into multiple spectral channels (first, second, and third spectral distributions) with different exposure times. Each channel captures spatial frequency components at different noise levels, allowing the system to process images in segments rather than requiring a single long exposure that would reduce throughput.
Solution Approach 2:
The system changes the exposure time parameter for different spectral channels - the first and second image planes use longer exposure times to capture low spatial frequency components with lower noise, while the third image plane uses a shorter exposure time for high spatial frequency components. This parameter variation allows maintaining high throughput while reducing overall image noise.
2Productivity
If the throughput of a document scanner is increased, then the productivity is improved, but the manufacturing cost increases due to requirement for better quality optics
Solution Approach 1:
The system introduces an intermediary processing stage that captures multiple spectral images with different exposure times and combines them through spatial frequency decomposition. This intermediary approach allows using standard optics rather than high-quality optics, reducing manufacturing costs while maintaining high throughput capability.
Solution Approach 2:
The final image is constructed as a composite from multiple spectral channels with different exposure characteristics. By combining the low-frequency components from longer exposure channels with high-frequency components from shorter exposure channels, the system achieves high-quality images using standard optics rather than requiring expensive high-quality optics throughout.
3Reliability
If a color transform is applied to all spatial frequency components, then the noise reduction is improved, but the high-frequency components become noisy
Solution Approach 1:
The color transform is applied selectively to different spatial frequency components based on their local characteristics. Low spatial frequency components undergo the full color transform for maximum noise reduction, while high spatial frequency components are preserved with minimal processing to maintain their inherent low-noise quality. This localized approach optimizes noise reduction without degrading high-frequency detail.
Solution Approach 2:
The color transform is applied partially - only to the first, second, and third transformed spatial frequency image planes corresponding to low-frequency components, rather than to all spatial frequency components. This partial application achieves sufficient noise reduction for the majority of the image while preserving the quality of high-frequency components that would otherwise become noisy.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method effectively reduces noise and enhances sensitivity, maintaining image quality while minimizing costs by applying the color transform only to low spatial frequency components and combining them with high-frequency components, which are less noisy.
Implementation Method 1
Image capture is the process of obtaining a digital image from a scene or a hard copy image such as an image printed on paper. This involves detecting light reflected from, or transmitted by or through the object of which the image is to be obtained.
Data Source
AI summary
First, second, and third image planes are obtained from at least one image sensor. The first image plane is formed from light of a first spectral distribution. The second image plane is formed from light of a second spectral distribution. The third image plane is formed from light of a spectral distribution which substantially covers the visible spectrum. First spatial frequency components are generated from the first, second and third image planes. A second spatial frequency component is generated from the third image plane. A color transform is applied to the first spatial frequency components from the first, second, and third image planes to obtain at least first, second and third transformed first spatial frequency image planes. The at least first, second and third transformed first spatial frequency image planes are combined with the second spatial frequency component from the third image plane to form an image.