Frequency Transform Image Noise Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting and measuring streaks in images are inadequate as they rely on edge detection algorithms sensitive to threshold settings, prone to noise, and lack directional information, leading to inaccurate severity measurements and potential false failures.
Innovation Solution
Applying a frequency transform to images to obtain a magnitude image, comparing threshold values, and reconstructing the image to remove noise, thereby improving noise detection and removal by identifying and setting undesired frequencies to zero.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If a low threshold is used in edge detection algorithm, then more pixels are detected, but the results become increasingly susceptible to noise
Solution Approach 1:
The patent transforms the image from spatial domain to frequency domain using Fourier transform. This dimensional change allows noise and streaks to be separated based on their frequency characteristics rather than spatial proximity, enabling selective removal of periodic noise without affecting edge detection sensitivity.
Solution Approach 2:
The patent applies different processing treatments to different frequency components in the frequency domain. By identifying specific frequency ranges corresponding to streaks and applying thresholding only to those regions, the method removes noise selectively while preserving important image features like edges.
2Object-affected harmful factors
If a high threshold is used in edge detection algorithm, then noise susceptibility is reduced, but subtle pixels which constitute a streak may be missed
Solution Approach 1:
By moving to frequency domain analysis, the patent enables precise measurement of streak severity through magnitude image analysis without being constrained by threshold selection in spatial domain. The frequency transform converts subtle variations into measurable magnitude differences.
Solution Approach 2:
The patent replaces the gradient magnitude calculation mechanism with a frequency transform-based magnitude analysis. This substitution provides a more robust measurement approach that doesn't rely on threshold selection and directional derivative calculations.
3Difficulty of detecting and measuring
If edge detection algorithm is used, then most edges can be detected, but directional information of the streak is not provided
Solution Approach 1:
The Fourier transform inherently preserves directional information in the frequency domain. By analyzing the phase and magnitude components, the patent can determine streak orientation and direction, which is lost in conventional edge detection methods.
4Measurement precision
If conventional defect density test is used, then localized densities can be measured, but the method allows possibilities of false failure due to lens roll-off or edge effects
Solution Approach 1:
By transforming to frequency domain, the patent converts spatial edge effects and lens roll-off into distinct frequency components that can be identified and excluded from defect analysis. This separates true defects from artifacts caused by optical system characteristics.
Data Source
AI summary
A method for detecting and removing noise from an image, the method includes the steps of applying a frequency transform to the image for obtaining a magnitude image; comparing a threshold value to values of the magnitude image for obtaining thresholded values; setting one or more thresholded values to a predetermined value; and reconstructing the image from the magnitude image having the predetermined values for removing noise from the image.