Q2. What future works have the authors mentioned in the paper "Robust rician noise estimation for mr images" ?
In future work, the proposed approach should be adapted to multi-channels signal acquisitions ( noncentral χ-distribution ) [ 4,21 ] and for images with non stationary noise such as those attributed to parallel imaging ( i. e. GRAPPA or SENSE ). The multi-channels signal acquisitions are becoming more often used during clinical acquisitions and the investigation of methods dedicated to these images should be further investigated.
Q3. Why did the authors choose to use classical WT in the proposed method?
Due to the additional computational burden and memory requirement needed by DTWT compared to WT, especially in 3D, the authors have chosen to use classical WT in the proposed method.
Q4. What is the capacity of the wavelet transforms to distinguish between noise and structure?
The capacity of the wavelet transforms to distinguish between noise and structure has been used in denoising methods to remove or reduce the coefficients corresponding to the noise components over the detailed sub-bands [14,15].
Q5. What is the main reason why the proposed method is being used in clinical acquisitions?
The multi-channels signal acquisitions are becoming more oftenused during clinical acquisitions and the investigation of methods dedicated to these images should be further investigated.
Q6. What is the main reason why the proposed method is overestimated?
The ML method proposed by Sijbers et al. tends to overestimate the noise variance as the noise power increases or when ghosting artefacts are added.
Q7. What is the common method to measure the noise variance in MR images?
A common manner to measure the Rician noise variance in magnitude MR images with large enough background areas is to estimate it from the mode of the histogram [36–38].
Q8. What is the definition of noise in MR?
In MR image analysis, the estimation of the noise level in an image is a mandatory step that must be addressed to assess the quality of the analysis and the consistency of the image processing technique.
Q9. What is the first method proposed by Sijbers et al.?
The first is a method proposed by Sijbers et al. [37] which is based on the maximum likelihood estimation principle over a partial histogram Hp.
Q10. What is the resulting iterative correction scheme?
The resulting iterative correction scheme can be written as:θt = √ ξ(θt−1) ( 1 +mo σ̂) − 2 (9)where mo is the mean signal of the object and σ̂ the first estimation from MAD estimator.