Bootstrap Resampling for Image Registration Uncertainty Estimation Without Ground Truth
read more
Citations
A Survey on Deep Learning-Driven Remote Sensing Image Scene Understanding: Scene Classification, Scene Retrieval and Scene-Guided Object Detection
Adjusted Fireworks Algorithm Applied to Retinal Image Registration
Summarizing and visualizing uncertainty in non-rigid registration
Probabilistic inference of regularisation in non-rigid registration.
Deformable image registration by combining uncertainty estimates from supervoxel belief propagation
References
An introduction to the bootstrap
Bootstrap Methods: Another Look at the Jackknife
Image registration methods: a survey
Related Papers (5)
Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI registration.
Frequently Asked Questions (13)
Q2. What is the criterion used to calculate the Hessian matrix?
In their implementation, image is interpolated using cubic B-splines [16], [17], its derivative is calculated analytically, and the minimization (4) is performed using the BFGS (Broyden-Fletcher-Goldfarb-Shanno) pseudoNewton algorithm [18], which incrementally updates the estimate of the Hessian matrix from the gradient.
Q3. What is the method for assessing the accuracy of a parametric estimator?
Bootstrap resampling [9]–[11], [47]–[50] is a powerful and versatile computational technique for assessing the accuracy of a parametric estimator in small sample situations.
Q4. What is the value of a bootstrap resampling?
Bootstrap resampling was used in image processing to evaluate the performance of detection and classification algorithms [51], [52] and edge detectors [53], to compensate the bias in estimation of ellipse parameters [54] and to improve image segmentation [55], [56].
Q5. What is the criterion for the hessian matrix?
Since images , are random (across realizations) due to the stochastic nature of the image generation process (measurement noise), the criterion is also random, and, hence, the estimate from (4) is random, too.
Q6. What is the effect of the trimmed mean?
To eliminate the influence of outliers (the optimization program failing to converge) and thus distorting the statistics, the authors used a trimmed mean, discarding of the highest and lowest values.
Q7. What is the coefficient of variation of the bootstrap method?
The authors observe that the coefficient of variation decreases with but the decrease is slow and diminishes even further with increased noise level .
Q8. What is the method for estimating the accuracy of a pixel?
Although for the sake of simplicity the authors have considered only 2-D translations, the presented accuracy estimation techniques are directly usable for other registration methods that find transformation with more degrees of freedom.
Q9. What is the log-likelihood of the Fisher information matrix?
The corresponding log-likelihood is(8)The elements of the Fisher information matrix (FIM) are(9)The second quadratic term in (8) is constant with respect to and the expected value of is zero.
Q10. What is the definition of the mean displacement variance?
In particular, the authors shall evaluate the covariance matrixwith (5)and a mean displacement variance(6)For , the expression simplifies to(7)The mean displacement variance is equal to the mean squared geometric error (MSE) provided that the estimator (4) is unbiased, .
Q11. What is the method for estimating SNR?
For medium to high SNR and Gaussian noise, the FRAE method (Section I-E) gives usable estimates that correctly follow the trend of the true error, even though the error is often overestimated [Fig. 2(a)–(f)].
Q12. How can the authors estimate the accuracy of the registration method?
The registration accuracy can also be estimated indirectly, from ground truth segmentations [28], [29] or by its ability tocreate good generative models [30].
Q13. What is the corresponding log-likelihood of the partial derivatives?
Hence(10)and using the chain rule yields(11)In accordance with [13], the authors estimate the partial derivatives using first order differences.