Pixon Map Iterative Update for Image Reconstruction Artifacts

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current pixon methods for image reconstruction require computing an initial pseudoimage and determining the pixon map after image update, which can introduce artifacts and are less effective in nonlocal transformations, such as those found in interferometry and tomography.

Innovation Solution

The pixon map is computed and updated during the iteration process using an updating variable, such as a gradient of a merit function, and smoothed with pixon kernels to select the widest kernel that meets a predetermined criteria, allowing for iterative refinement of the image reconstruction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the pixon map is determined after computing an initial pseudoimage, then the reconstruction can proceed with a defined pixon map, but artifacts are introduced and the method is less effective for nonlocal transformations

Engineering Contradiction:
Improvereconstruction accuracyVSAvoidartifacts
Core Design Contradiction:
ReliabilityVSObject-generated harmful factors

Solution Approach 1:

The pixon map is determined in advance from the updated image before the smoothing step is applied. This preliminary determination ensures that the pixon map reflects the actual image structure rather than being based on a pseudoimage, thereby preventing artifact introduction while maintaining reconstruction accuracy.

Inventive Principle:
Principle #10Preliminary action

2Object-generated harmful factors

If the pixon map is determined from the updated image, then artifacts are reduced, but the method requires additional computation of the updated image first

Engineering Contradiction:
ImproveartifactsVSAvoidcomputation efficiency
Core Design Contradiction:
Object-generated harmful factorsVSProductivity

Solution Approach 1:

The determination of the pixon map is merged with the existing image update step in the iterative reconstruction algorithm. Both the updated image and the pixon map are computed from the same updated image data, eliminating redundant computations and maintaining efficiency while improving accuracy.

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If standard iterative methods are used to converge to maximum-likelihood solution, then the solution is statistically optimal, but convergence is slow even when terminated early to avoid overfitting

Engineering Contradiction:
Improvestatistical optimalityVSAvoidconvergence time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The pixon map provides adaptive feedback during the iterative reconstruction process. By determining the pixon map from the updated image at each iteration, the algorithm dynamically adjusts the smoothing based on the current image state, accelerating convergence while preventing overfitting through early termination.

Inventive Principle:
Principle #23Feedback

4Loss of time

If the Hessian matrix is used to achieve faster convergence, then convergence speed improves, but the matrix is too large to be computed or stored for large-scale problems

Engineering Contradiction:
Improveconvergence timeVSAvoidcomputational resources
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

Instead of using a global Hessian matrix that requires storing all second-order derivatives, the pixon map provides local adaptive smoothing at each pixel position. This local approach achieves faster convergence without requiring the computation or storage of a large global matrix, making it feasible for large-scale problems.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP2697740B1Method to determine a pixon map in iterative image reconstruction
Publication Date: 2018.08.15 SIEMENS MEDICAL SOLUTIONS USA INC
  • EP2697740B1 patent drawingFigure 1
  • EP2697740B1 patent drawingFigure 2~3
  • EP2697740B1 patent drawingFigure 4~5

AI summary

A method for iterative reconstruction of a signal containing noise using the pixon method determines the pixon map from a variable that is used to update the image in the iteration. The updating variable is based on an optimized merit function, and smoothes the updating variable during the iteration. The updated image can optionally also be further smoothed at the end of the iteration, using the pixon map determined during the iteration.