Skip to main content
Log in

Regularized restoration for two dimensional band-limited signals

  • Published:
Multidimensional Systems and Signal Processing Aims and scope Submit manuscript

Abstract

In this paper the ill-posedness of restoring lost samples is discussed in the two dimensional case. The restoration algorithm by Shannon’s Sampling Theorem is analyzed. A regularized restoring algorithm for two dimensional band-limited signals is presented. The convergence of the regularized restoring algorithm is studied and compared with the restoration algorithm by Shannon’s sampling theorem in the two dimensional case.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1

Similar content being viewed by others

References

  • Belge, M., Kilmer, M. E., & Miller, E. L. (2002). Efficient determination of multiple regularization parameters in a generalized L-curve framework. Inverse Problems, 18, 1161–1183.

    Article  MATH  MathSciNet  Google Scholar 

  • Chen, W. (2006). An efficient method for an Ill-posed problem–band-limited extrapolation by regularization. IEEE Transactions on Signal Processing, 54, 4611–4618.

    Article  Google Scholar 

  • Chen, W. (2011). The Ill-posedness of restoring lost samples and regularized restoration for band-limited signals. Signal Processing, 91, 1315–1318.

    Article  MATH  Google Scholar 

  • Cheung, K. F., & Marks, II R. J. (May 1985). Ill-posed sampling theorems. IEEE Transactions on Circuits and Systems, vol. CAS-32, 481–484.

  • Eldar, Y. C., & Unser, M. (2006). Nonideal sampling and interpolation from noisy observations in shift-invariant spaces. IEEE Transactions on Signal Processing, 54(7).

  • Griesbaum, A., Barbara, B., & Vexler, B. (2008). Efficient computation of the Tikhonov regularization parameter by goal-oriented adaptive discretization. Inverse Problems, 24, 1–20.

    Article  Google Scholar 

  • Kilmer, M. E., & O’leary, D. P. (2001). Choosing regularization parameters in iterative methods for ill-posed problems. SIAM Journal on Matrix Analysis and Applications, 22(4), 1204–1221.

    Article  MATH  MathSciNet  Google Scholar 

  • Marks II, R. J. (1983). Restoring lost samples from oversampled band-limited signal. IEEE Transactions on Signal Processing, ASSP-31, 752–755.

  • Marks, R. J, I. I. (1991). Introduction to shannon sampling and interpolation theory. New York: Springer.

    Book  MATH  Google Scholar 

  • Selva, J. (2009). Functionally weighted Lagrange interpolation of band-limited signals from nonuniform samples. IEEE Transactions on Signal Processing, 57, 168–181.

    Article  MathSciNet  Google Scholar 

  • Shannon, C. E. (July 1948). A mathematical theory of communication. The Bell System Technical Journal, 27, 379–423.

  • Smale, S., & Zhou, D.-X. (2004). Shannon sampling and function reconstruction from point values. Journal: Bulletin of the American Mathematical Society, 41, 279–305.

    MATH  MathSciNet  Google Scholar 

  • Tikhonov, A. N., & Arsenin, V. Y. (1977). Solution of Ill-posed problems. Winston: Wiley.

    Google Scholar 

Download references

Acknowledgments

The author would like to express appreciation to professor Malcolm R. Adams for his help in the revision of this paper.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Weidong Chen.

Appendix

Appendix

Proof of Theorem 1.2

In (1) we take the convolution with

$$\begin{aligned} \frac{\sin \Omega _0t_1}{\Omega _0t_1} \frac{\sin \Omega _0t_2}{\Omega _0t_2} \end{aligned}$$

we obtain

$$\begin{aligned} f(t_1, t_2)=r_1r_2\sum _{n_1=-\infty }^\infty ~\sum _{n_2=-\infty }^\infty \frac{\sin \Omega _0(t_1-n_1h_1)}{\Omega _0(t_1-n_1h_1)} \frac{\sin \Omega _0(t_2-n_2h_2)}{\Omega _0(t_2-n_2h_2)} f(n_1h_1, n_2h_2). \end{aligned}$$

Let \(t_1=0, ~t_2=0\), solve for \(f(0, 0)\), then we obtain the formula (2).

Proof of Lemma 3.3

By the Cauchy inequality and Lemma 3.2

$$\begin{aligned} S_1^2\le \sum _{n_1\not =0} \left| \frac{[2\pi \alpha +2\pi \alpha (n_1h_1)^2]}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2] n_1h_1}\right| ^2 \left( \frac{h_1}{r_1\pi }\right) ^2 \sum _{n_1\not =0} |f(n_1h_1, 0)|^2 =O\left( \alpha ^{\frac{1}{2}}\right) . \end{aligned}$$

Proof of Lemma 3.4

$$\begin{aligned} S_2&= \sum _{n_1\not =0}~\sum _{ n_2=0} \frac{\eta (n_1h_1, n_2h_2)}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]}\frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi } \\&+\sum _{n_1=0}~\sum _{n_2\not =0} \frac{\eta (n_1h_1, n_2h_2)}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]}\frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi }\\&+\sum _{n_1\not =0}~\sum _{n_2\not =0} \frac{\eta (n_1h_1, n_2h_2)}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]}\frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi } \end{aligned}$$

By Lemma 3.1, the first two terms are of the order \(O(\delta )+O(\delta \ln \frac{1}{\alpha })\). By using the estimation (6) the third term is of the order \(O(\delta \ln ^2 \frac{1}{\alpha })\).

Proof of Lemma 3.5

By the Cauchy inequality

$$\begin{aligned}&I^2\le \sum _{n_1\not =0} ~\sum _{n_2\not =0} \left| \frac{2\pi \alpha + 2\pi \alpha (n_1h_1)^2}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2] n_1h_1} \frac{2\pi \alpha + 2\pi \alpha (n_2h_2)^2}{[1+2\pi \alpha +2\pi \alpha (n_2h_2)^2] n_2h_2}\right| ^2 \\&\cdot \left( \frac{h_1}{r_1\pi }\right) ^2 \left( \frac{h_2}{r_2\pi }\right) ^2 \sum _{n_1\not =0} ~\sum _{n_2\not =0} |f(n_1h_1, n_2h_2)|^2. \end{aligned}$$

Then by Lemma 3.2, \(I^2=O(\alpha )\).

Proof of Lemma 3.6

By the Cauchy inequality

$$\begin{aligned}&J_1^2\le \sum _{n_1\not =0} ~\sum _{n_2\not =0} \left| \frac{2\pi \alpha + 2\pi \alpha (n_1h_1)^2}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2] n_1h_1} \frac{1}{[1+2\pi \alpha +2\pi \alpha (n_2h_2)^2] n_2h_2} \right| ^2 \\&\cdot \left( \frac{h_1}{r_1\pi }\right) ^2 \left( \frac{h_2}{r_2\pi }\right) ^2\sum _{n_1\not =0} \sum _{n_2\not =0} |f(n_1h_1, n_2h_2)|^2. \end{aligned}$$

Then by Lemma 3.2, \(J_1^2=O(\alpha ^\frac{1}{2})\).

Proof of Lemma 3.7

\(J_2=I_1+I_2+I_3\) where

$$\begin{aligned} I_1&= \sum _{n_1\not =0}~\sum _{n_2=0} \frac{\left\{ [1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]-1\right\} }{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]} \\&\cdot \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi } f(n_1h_1, n_2h_2)\\ I_2&= \sum _{n_1=0}~\sum _{n_2\not =0} \frac{\left\{ [1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]-1\right\} }{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]}\\&\cdot \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi } f(n_1h_1, n_2h_2)\\ I_3&= \sum _{n_1\not =0}~\sum _{n_2\not =0} \frac{\left\{ [1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]-1\right\} }{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]} \\&\cdot \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi } f(n_1h_1, n_2h_2) \end{aligned}$$

Here

$$\begin{aligned} I_1&= \sum _{n_1\not =0} \frac{\left\{ [1+2\pi \alpha +2\pi \alpha (n_1h_1)^2](1+2\pi \alpha )-1\right\} }{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2](1+2\pi \alpha )} \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } f(n_1h_1, 0)\\&= \frac{1}{1+2\pi \alpha } \sum _{n_1\not =0} \frac{[2\pi \alpha +2\pi \alpha (n_1h_1)^2]}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2]} \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } f(n_1h_1, 0)\\&+\frac{2\pi \alpha }{1+2\pi \alpha } \sum _{n_1\not =0} \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } f(n_1h_1, 0) =O(\alpha ^{\frac{1}{4}}) \end{aligned}$$

by Lemma 3.3. For the same reason \(I_2=O(\alpha ^{\frac{1}{4}})\).

$$\begin{aligned} I_3&= \sum _{n_1\not =0}~\sum _{n_2\not =0} \frac{[2\pi \alpha +2\pi \alpha (n_1h_1)^2][2\pi \alpha +2\pi \alpha (n_2h_2)^2]}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]} \\&\cdot \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi } f(n_1h_1, n_2h_2)\\&+\sum _{n_1\not =0}~\sum _{n_2\not =0} \frac{[2\pi \alpha +2\pi \alpha (n_1h_1)^2]}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]}\\&\cdot \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi } f(n_1h_1, n_2h_2)\\&+\sum _{n_1\not =0}~\sum _{n_2\not =0} \frac{[2\pi \alpha +2\pi \alpha (n_2h_2)^2]}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]}\\&\cdot \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi } f(n_1h_1, n_2h_2) =O(\alpha ^{\frac{1}{4}}) \end{aligned}$$

by Lemma 3.5 and by Lemma 3.6.

Proof of Theorem 3.1

\(f_\alpha (0, 0)-f_E(0, 0)\)

$$\begin{aligned}&= \frac{r_1r_2}{1-r_1r_2}\sum _{n_1\not =0}~\sum _{\mathrm{or} ~n_2\not =0} \frac{f(n_1h_1, n_2h_2)+\eta (n_1h_1, n_2h_2)}{[1+2\pi \alpha +2\pi \alpha (n_1h_1)^2][1+2\pi \alpha +2\pi \alpha (n_2h_2)^2]} \\&\cdot \frac{\sin (n_1r_1\pi )}{n_1r_1\pi } \frac{\sin (n_2r_2\pi )}{n_2r_2\pi } \\&\quad -\frac{r_1r_2}{1-r_1r_2} \sum _{n_1\not = 0} ~\sum _{\mathrm{or} ~n_2\not =0} \frac{\sin (n_1r_1\pi )}{n_1r_1\pi }\frac{\sin (n_2r_2\pi )}{n_2r_2\pi } f(n_1h_1, n_2h_2) =\frac{r_1r_2}{1-r_1r_2}(J_2+S_2) \end{aligned}$$

where \(J_2=O(\alpha ^{\frac{1}{4}})\) is the same as in Lemma 3.7 and

$$\begin{aligned} S_2=O(\delta )+O\left( \delta \ln \frac{1}{\alpha }\right) +O\left( \delta \ln ^2 \frac{1}{\alpha }\right) \end{aligned}$$

is the same as in Lemma 3.4.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Chen, W. Regularized restoration for two dimensional band-limited signals. Multidim Syst Sign Process 26, 665–675 (2015). https://doi.org/10.1007/s11045-013-0263-2

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11045-013-0263-2

Keywords

Navigation