Radiograph Generation Weighting Factor Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radiograph generation methods suffer from non-uniform signal-to-noise ratios due to interpolation methods during backprojection, leading to noise and artifact issues, particularly in regions with high-intensity differences and non-conformities, which are exacerbated by methods like bilinear and nearest neighbor interpolation.
Innovation Solution
A radiograph generation method that sets weighting factors for combining discrete image processing algorithms, such as nearest neighbor and bilinear interpolation, based on pixel value differences to minimize negative effects and maximize positive effects, thereby reducing noise and artifacts through weight addition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Shape
If bilinear interpolation is used during backprojection, then image smoothness is improved, but non-uniform signal-to-noise ratio and grid artifacts occur
Solution Approach 1:
The patent applies different interpolation methods to different regions of the image based on local characteristics. Specifically, it identifies regions with high signal-to-noise ratio and applies bilinear interpolation there, while using nearest neighbor interpolation in regions with low signal-to-noise ratio or at pixel value edges, thereby optimizing image quality locally rather than uniformly across the entire image
Solution Approach 2:
The patent dynamically selects the interpolation method based on local image characteristics such as signal-to-noise ratio and pixel value gradients. The interpolation approach changes adaptively across different regions rather than being static, allowing the system to optimize between smoothness and artifact reduction based on local conditions
2Productivity
If nearest neighbor interpolation is used, then calculation cost is reduced, but unevenness occurs at pixel value edges
Solution Approach 1:
The patent applies nearest neighbor interpolation specifically in regions where it performs well (low signal-to-noise ratio areas) while using bilinear interpolation in regions where smoothness is more critical (high signal-to-noise ratio areas or regions away from edges), thereby optimizing both calculation efficiency and image quality in different local contexts
3Object-affected harmful factors
If Lanczos interpolation is used, then signal-to-noise uniformity is improved, but calculation cost and processing time increase
Solution Approach 1:
Instead of applying the computationally expensive Lanczos interpolation uniformly across the entire image, the patent applies it only partially in specific regions where it provides the most benefit, while using simpler interpolation methods in other regions, thereby achieving signal-to-noise uniformity improvement without the full calculation cost
Solution Approach 2:
The patent changes the interpolation parameter (method selection) based on local image characteristics such as signal-to-noise ratio and position relative to pixel value edges, transitioning between different interpolation approaches rather than using a single fixed method, thereby optimizing the balance between quality and efficiency
Data Source
AI summary
An image processing element 9 of the present invention includes a weighting factor setting element 9a that sets up a weighting factors wnn, wbi of the weight addition combing a nearest neighbor interpolation and a bilinear interpolation based on an absolute value |Ibi−Inn| of the difference between the pixel value Inn acquired by the nearest neighbor interpolation and the pixel value Ibi acquired by the bilinear interpolation; and a weight addition element 9b that implement a weighting addition based on the set-up weighting factors wnn, wbi. A reconstructed image can be acquired by arranging the backprojection pixel value Inew every pixel following the weighting. A reconstructed image can be obtained in which any noise and artifacts are reduced by setting the weighting factor of the image processing algorithm causing a negative effect on the region of some pixel values to be minimal and setting the weighting factor of the (other) image processing algorithm causing a positive-effect on the same pixel value region to be maximum based on the absolute value |Ibi−Inn| of the difference.