Medical Image Noise Filtering via Weighted Graph Dirichlet Boundary
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for filtering noise in medical images, such as those by Perona and Malik, fail to account for inhomogeneous data sampling and lack a method for choosing convergence criteria, leading to inadequate noise reduction and preservation of inter-object contrast.
Innovation Solution
A system and method that generates a weighted graph representing the medical image, selects nodes to retain grayscale values, and determines filtered values by solving a combinatorial anisotropic Dirichlet boundary value problem, allowing for flexible noise reduction and contrast preservation through Dirichlet boundary conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If nonlinear diffusion methods are used to reduce image noise, then noise is reduced, but inter-object contrast is lost and convergence criteria cannot be determined
Solution Approach 1:
The patent applies different diffusion behaviors to different regions of the image by using inhomogeneous sampling densities. Regions with higher sampling density undergo more diffusion iterations, while regions with lower sampling density undergo fewer iterations, allowing local adaptation that preserves edges while reducing noise in homogeneous areas
Solution Approach 2:
The patent dynamically adjusts the number of diffusion iterations based on local sampling density rather than applying a fixed stopping parameter globally. This dynamic approach allows the filtering process to adapt to local image characteristics, preserving contrast at edges while effectively reducing noise in uniform regions
2Object-affected harmful factors
If Perona-Malik anisotropic diffusion is applied, then noise filtering is achieved, but the method lacks adaptability for inhomogeneous data sampling and tuning constants
Solution Approach 1:
The patent makes the diffusion process adaptive to local sampling conditions by using inhomogeneous sampling densities that reflect local image characteristics. This allows the method to automatically adapt to different regions without requiring manual tuning of constants for each region
Solution Approach 2:
The method uses the image's own sampling density distribution to guide the filtering process. Regions with different sampling densities automatically receive appropriate amounts of filtering based on their local characteristics, making the system self-adapting without external intervention
3Loss of time
If a stopping parameter is used to control diffusion time, then the process can be terminated, but there is no method for choosing appropriate convergence criteria
Solution Approach 1:
The patent eliminates the need for manual selection of stopping parameters by using inhomogeneous sampling density as an intrinsic convergence criterion. The sampling density distribution itself determines where and how much filtering should occur, making the method self-regulating and easy to operate without expert knowledge of diffusion parameters
Data Source
AI summary
A system and method for filtering noise from a medical image are provided. A method for filtering noise from an image comprises: generating a weighted graph representing the image; selecting a plurality of nodes from the image to retain grayscale values of the plurality of nodes; and determining filtered grayscale values of the plurality of nodes.