Linear Noise Estimation in Image Processing for Quantitative Fidelity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques that reduce linear noise in images lose quantitativeness after noise reduction processing.
Innovation Solution
An image processing apparatus and method that estimates a noise image using an evaluation function incorporating differentiation and low-frequency component extraction processing to minimize noise while maintaining quantitativeness, employing a noise estimation unit and noise reduction unit to generate a noise-reduced image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If noise reduction processing is applied to remove linear noises from the target image, then the image quality is improved, but the quantitativeness of the image is lost
Solution Approach 1:
The image is segmented into signal components and noise components through iterative processing. The algorithm separates linear noises from the actual image data by repeatedly applying differentiation and low-pass filtering operations, allowing selective removal of noise while preserving the quantitativeness of the original signal.
Solution Approach 2:
The noise reduction process employs periodic iterative operations where the target image is repeatedly differentiated, low-pass filtered, and compared with the original image. This cyclic process continues until convergence, systematically removing linear noises while maintaining the quantitative integrity of the underlying signal.
2Measurement precision
If differentiation processing and low frequency component extraction are applied to estimate noise, then noise estimation accuracy is improved, but the processing complexity increases
Solution Approach 1:
The patent replaces complex mechanical or manual noise analysis with automated computational operations. Differentiation and low-pass filtering are implemented through standard digital signal processing algorithms, substituting manual image analysis with systematic mathematical operations that can be efficiently executed by computers.
Solution Approach 2:
The algorithm creates multiple copies of the target image through iterative differentiation and filtering operations. Each iteration generates a processed version that is compared with the original, allowing the system to estimate noise characteristics without directly modifying the original quantitative data until the final reconstruction stage.
Data Source
AI summary
In a noise estimation step, a noise image included in a target image including linear noises extending along a first direction is estimated. In this case, an evaluation function including a first term representing a difference between a result obtained by performing differentiation processing in a second direction and low frequency component extraction processing in the first direction on the target image and a result obtained by performing differentiation processing in the second direction and low frequency component extraction processing in the first direction on the noise image is used to obtain the noise image which minimizes a value of the evaluation function. In a noise reduction step, a noise reduced image is generated from the target image based on the target image and the noise image.