Medical Image Noise Reduction via Rotated Wavelet Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Noise reduction in medical diagnostic images, such as nuclear medicine and MRI images, is inadequate due to varying signal-to-noise ratios and large pixel sizes, leading to overcorrection and artifact generation in existing high-frequency cutoff filtering methods.
Innovation Solution
An image processing apparatus that generates rotated digital images, applies wavelet transform and local cutoff frequency calculations to determine specific filtering for each region, and combines processed images to achieve uniform noise reduction and eliminate artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the same high-frequency cutoff processing is performed through the whole image, then noise reduction is achieved, but overcorrection and insufficient correction occur due to local S/N ratio variations
Solution Approach 1:
The patent divides the image into multiple local regions and calculates a specific cutoff frequency for each region based on its local S/N ratio characteristics. This allows each region to receive customized filtering treatment, preventing overcorrection in high S/N regions and insufficient correction in low S/N regions, thereby maintaining both noise reduction effectiveness and spatial resolution/contrast quality.
2Reliability
If Fourier transform filtering is performed on images with large pixel size, then noise reduction is achieved, but artifacts are generated due to insufficient sampling
Solution Approach 1:
The patent segments the image processing into multiple rotations and applies filtering in the rotational domain rather than directly in the spatial domain. By processing multiple rotated versions of the image and combining the results, the method effectively increases sampling density and eliminates artifacts caused by large pixel sizes while maintaining noise reduction benefits.
3Reliability
If high-frequency cutoff filtering is applied to medical diagnostic images, then noise is reduced, but information deterioration occurs due to local S/N ratio variations
Solution Approach 1:
The patent dynamically changes the cutoff frequency parameter for each local region based on the measured S/N ratio characteristics of that region. By adapting the filtering parameter to local conditions rather than using a fixed global parameter, the method preserves signal information in high S/N regions while effectively reducing noise in low S/N regions, thereby minimizing information loss.
Data Source
AI summary
An image processing apparatus includes a storing section which stores data of a digital image, a rotation processing section which generates a plurality of rotated digital images having different rotation angles from the digital image, an image processing section which generates a plurality of image-processed digital images from the rotated digital images, a reverse processing section which generates a plurality of reversed digital images from the image-processed digital images, and a combining section which combines the reversed digital images into one digital image.