ForWaRD: Fourier-wavelet regularized deconvolution for ill-conditioned systems
read more
Citations
Going deeper with convolutions
Removing camera shake from a single photograph
A New Alternating Minimization Algorithm for Total Variation Image Reconstruction
A New TwIST: Two-Step Iterative Shrinkage/Thresholding Algorithms for Image Restoration
High-quality motion deblurring from a single image
References
A wavelet tour of signal processing
De-noising by soft-thresholding
Estimation with Quadratic Loss
Complex Wavelets for Shift Invariant Analysis and Filtering of Signals
Related Papers (5)
Frequently Asked Questions (11)
Q2. What are the future works mentioned in the paper "Forward: fourier-wavelet regularized deconvolution for ill-conditioned systems" ?
There are several avenues for future ForWaRD-related research. Construction of a universally applicable deconvolution scheme lying between WVD and VWD appears promising but challenging.
Q3. How can the authors compute the wavelet coefficients of discrete-time signals?
For a discrete-time signal with samples, the wavelet coefficients can be efficiently computed in operations using a filterbank consisting of lowpass filters, highpass filters, upsamplers, and decimators [7].
Q4. What is the motivation for the hybrid approach?
The motivation for the hybrid approach stems from the realization that deconvolution techniques relying on scalar shrinkage in a single transform domain—for example, the LTI Wiener deconvolution filter or the WVD—are inadequate to handle the wide variety of practically encountered deconvolution problems.
Q5. What is the sampling error for the range s > 1=p?
if the sampling kernel of (7) is employed, then the interpolation error is negligible with respect to the estimation error for the range s > 1=p 1=2; the range decreases to s > 1=p if impulse sampling is employed [19], [20].
Q6. How does the hybrid ForWaRD algorithm estimate from in (2)?
The hybrid ForWaRD algorithm estimates from in (2) by employing scalar shrinkage both in the Fourier domain to exploit its economical colored noise representation and in the wavelet domain to exploit its economical signal representation.
Q7. What is the optimal scaleregularization parameter for the Fourier shrinkage?
Since the noise standard deviation is primarily determined by the Fourier structure of the convolution operator, the authors can infer that the balance between Fourier and wavelet shrinkage is simultaneously determined by the Fourier structure of the operator and the wavelet structure of the desired signal.
Q8. How can the authors make the approximation of using circular convolution more precise?
For infinite-support and , the approximation of using circular convolution can be made arbitrarily precise by increasing the period.
Q9. What is the condition for the wavelet coefficients of a continuous-time 1-D signal?
The wavelet coefficients computed from samples (refer Section II) of a continuous-time 1-D signal, , satisfy (for all )(15) assuming sufficiently smooth wavelet basis functions [5], [19], [20].
Q10. What is the simplest way to show that the Fourier distortion error decays as?
If the parameterizing is tuned such that(27)with(28)for some constant , then the per-sample ForWaRD MSE decays as(29)as with a constant.
Q11. Who was the associate editor coordinating the review of this paper?
The associate editor coordinating the review of this paper and approving it for publication was Dr. Chong-Yung Chi.R. Neelamani was with the Department of Electrical and Computer Engineering, Rice University, Houston, TX 77005-1892 USA.