MIP Image Rendering Chessboard Artifact Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Maximum Intensity Projection (MIP) images often suffer from chessboard artifacts due to wideband noise in medical imaging datasets, which existing methods fail to address effectively without compromising image resolution.
Innovation Solution
A method that involves shooting rays from every pixel through a 3D voxel grid, using chessboard mitigation algorithms to sample intensity values with defined step sizes, and computing weighting coefficients to minimize noise variance differences, while applying localized low pass filtering, high-order interpolation, and distance remapping to reduce artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If pre-filtering is applied to reduce wideband noise, then chessboard artifacts are reduced, but high frequency components are damaged and MIP image resolution is lowered
Solution Approach 1:
The patent applies preliminary action by pre-computing a lookup table of filtered ray samples during an offline processing stage. This lookup table is generated by filtering the volumetric data and storing the resulting ray samples, which are then used during online rendering to avoid applying filters in real-time. This resolves the contradiction by performing the filtering operation beforehand when computational resources are available, rather than during the critical rendering path.
Solution Approach 2:
The patent uses copying by creating a lookup table that stores pre-filtered ray samples. Instead of recomputing filtered samples during each rendering operation, the system copies the pre-computed filtered samples from the lookup table during rendering. This significantly reduces the computational burden during rendering while maintaining the artifact reduction benefits of filtering.
2Productivity
If standard MIP rendering is used, then processing is simple and fast, but chessboard artifacts appear in the output image
Solution Approach 1:
The patent performs the filtering operation as a preliminary action during an offline processing stage, before rendering. By pre-filtering the volumetric data and storing the results in a lookup table, the system eliminates the need for complex filtering operations during the rendering process, thus maintaining fast rendering speeds while reducing chessboard artifacts.
Solution Approach 2:
The patent employs copying by using a lookup table to store pre-computed filtered ray samples. During rendering, the system simply copies these pre-filtered samples from the lookup table rather than performing complex filtering operations in real-time. This approach maintains rendering speed while eliminating artifacts.
3Object-affected harmful factors
If filtering is applied during rendering, then artifacts are reduced, but computational complexity increases
Solution Approach 1:
The patent applies filtering as a preliminary action during an offline processing stage rather than during rendering. This separates the computationally intensive filtering operation from the rendering process, allowing the system to pre-compute filtered samples when computational resources are available and then simply retrieve them during rendering.
Solution Approach 2:
The patent uses copying by creating a lookup table that stores pre-filtered ray samples. During rendering, the system copies these pre-computed samples from the lookup table, which is a simple memory access operation compared to performing filtering in real-time. This significantly reduces computational complexity during rendering while maintaining artifact reduction.
Data Source
AI summary
A system and method for rendering an MIP image that has reduced high frequency component loss and reduced chessboard artifacts. The method includes accessing volumetric data, having random noise. Rays are shot, or cast, through the volumetric data, onto a voxel grid, which has grid points. Sampling data along each ray is performed to obtain selected sample data points on the ray and a distance from a selected point to a nearest grid point is determined. A voxel intensity value is accessed at each selected point, as a function of the position of the selected point relative to the nearest grid point. The difference between the voxel intensity at each selected point is minimized an image is rendered from the volumetric data as a function of the minimizing step. Multiple methods may be used to minimize the difference between the voxel intensity at each selected point. These include for example 1.) applying a localized low pass filter, 2.) applying a localized high order interpolation kernel, and 3.) applying localized distance remapping methods.