## Interpolation and the Discrete Papoulis-Gerchberg Algorithm (1994)

Venue: | IEEE Trans. Signal Processing |

Citations: | 32 - 20 self |

### BibTeX

@ARTICLE{Ferreira94interpolationand,

author = {Paulo Jorge S. G. Ferreira},

title = {Interpolation and the Discrete Papoulis-Gerchberg Algorithm},

journal = {IEEE Trans. Signal Processing},

year = {1994},

volume = {42},

pages = {2596--2606}

}

### OpenURL

### Abstract

In this paper we analyze the performance of an iterative algorithm, similar to the discrete Paponiis-Gerchberg algorithm, and which can be used to recover missing samples in finite-length records of band-limited data. No assumptions are made regarding the distribution of the missing samples, in contrast with the often studied extrapolation problem, in which the known samples are grouped together. Indeed, it is possible to regard the observed signal as a sampled version of the original one, and to interpret the reconstruction result studied herein as a sampling result. We show that the iterative algorithm converges if the density of the sampling set exceeds a certain minimum value which naturally increases with the bandwidth of the data. We give upper and lower bounds for the error as a function of the number of iterations, together with the signals for which the bounds are attained. Also, we analyze the effect of a relaxation constant present in the algorithm on the spectral radius of the iteration matrix. From this analysis we infer the optimum value of the relaxation constant. We also point out, among all sampling sets with the same density, those for which the convergence rate of the recovery algorithm is maximum or minimum. For low-pass signals it turns out that the best convergence rates result when the distances among the missing samples are a multiple of a certain integer. The worst convergence rates generally occur when the missing samples are contiguous.

### Citations

679 | Matrix Iterative Analysis - Varga - 1962 |

307 |
Prolate spheroidal wave functions, Fourier analysis and uncertainty
- Slepian, Pollak
- 1961
(Show Context)
Citation Context ... III. RELATED WORK A large percentage of the many published papers concerned with iterative constrained signal restoration is devoted to the band-limited extrapolation problem. Its study goes back to =-=[8]-=-, where an extrapolation method is given, based on the double orthogonality property of the prolate spheroidal wave functions. The introduction of the error-reduction algorithm, also known as the Papo... |

127 |
Image Restoration by the Method of Convex Projections: Part 1 - Theory
- Youla, Webb
- 1982
(Show Context)
Citation Context ...e projections onto convex sets may also outperform the Papoulis-Gerchberg iteration, since they allow any number of a priori constraints to be included in the restoration procedure. See, for example, =-=[23]-=--[26]. An iterative algorithm for restoring continuously sampled band-limited L2 functions was proposed in [27]. A continuously sampled signal is obtained by setting to zero a signal except for a neig... |

107 |
A new algorithm in spectral analysis and band-limited extrapolation
- Papoulis
- 1975
(Show Context)
Citation Context ...mples is known. The reconstruction algorithm studied herein is an itemtire method which reduces to the discrete finite-dimensional version of the well-known Papoulis-Gerchberg extrapolation algorithm =-=[2]-=-, [31, if the unknown samples are contiguous, and if a relaxation constant/ is unitary. In the latter case, the algorithm may also be described as an alternating projection algorithm of the type found... |

97 | Super-resolution through error energy reduction, Optica Acta 21 - Gerchberg - 1974 |

74 | Fixed Point Theorems - Smart - 1974 |

59 |
Generalized image restoration by the method of alternating orthogonal projections
- Youla
- 1978
(Show Context)
Citation Context ...1, if the unknown samples are contiguous, and if a relaxation constant/ is unitary. In the latter case, the algorithm may also be described as an alternating projection algorithm of the type found in =-=[4]-=-. Its convergence will be established under no restriction on the distribution of the missing samples, generalizing results found in [5] and providing a theoretical background for the observations rep... |

59 | On non-uniform sampling of bandwidth limited signals - Yen - 1956 |

45 |
Constrained Iterative Restoration Algorithms
- Schafer, Mersereau, et al.
- 1981
(Show Context)
Citation Context ...uences with finite energy. A similar approach, based on fixed-point theorems, can be outlined under rather general conditions and leads to the constrained iterative restoration framework described in =-=[39]-=-. The extrapolation problem for multidimensional finite energy sequences was studied in [14], and related to the L2 band-limited extrapolation problem. More precisely, it was shown that the discrete s... |

35 | Image restoration by the method of convex projections: Part 2-applications and numerical results - Sezan, Stark - 1982 |

32 |
A transformation method for the reconstruction of functions from nonuniformly spaced samples
- Clark, Palmer, et al.
(Show Context)
Citation Context ...d both one-dimensional and twodimensional reconstruction problems are studied. In fact, it is possible to arrive at sampling expansions for signals which are not band-limited if such a mapping exists =-=[46]-=-. An alternative, which sometimes leads to good results, is to interpolate a set of uniformly spaced points on the nonuniform grid [47]. The Papoulis-Gerchberg algorithm can also be used [48] to obtai... |

29 |
Recovery of signals from nonuniform samples using iterative methods
- Analoui, Gamshadzahi
- 1991
(Show Context)
Citation Context ...isGerchberg algorithm and to a general iteration proposed by Sandberg [28], as discussed in [29]-[311. It was shown recently 2598 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 42, NO. 10, OCTOBER 1994 =-=[32]-=- that the same procedure can be used in the case of instantaneous sampling. Many discrete extrapolation problems lead to ill-conditioned systems of linear equations. Their noniterative solution is att... |

27 | Extrapolation algorithms for discrete signals with application in spectrum estimation - Jain, Ranganath - 1981 |

24 |
Noniterative and faster iterative methods for interpolation and extrapolation
- Ferreira
- 1994
(Show Context)
Citation Context ..., [3], brought a new interest to the problem, and, since then, several works dealing with different aspects of the extrapolation problem have appeared. We refer the interested reader to [1], [7], [9]-=-=[19]-=- among many others. These works address topics that include implementation aspects, numerical stability, convergence acceleration, noniterative extrapolation, the sampled analog of the extrapolation p... |

21 |
Extrapolation and spectral estimation with iterative weighted norm modification
- Cabrera, Parks
- 1991
(Show Context)
Citation Context ...he finite dimensionality of the problem. This is an important fact since it allows some of the difficulties that arise when sampling the solution of a continuous problem to be overcome. More recently =-=[55]-=-, a finite-dimensional algorithm for interpolation and extrapolation was proposed which requires less a priori information about the bandwidth of the data, and which can be applied to mixed interpolat... |

19 |
Analysis of error in reconstruction of two-dimensional signals from irregularly spaced samples
- Chen, Allebach
- 1987
(Show Context)
Citation Context ...Although the result does not have minimum norm, it is optimum for signals concentrated on that interval. The estimation of two-dimensional signals from a finite number of samples was also studied. In =-=[52]-=- the authors choosed to minimize the maximum of the mean-squared error over a class of signals with respect to the sampling set. They found that the optimum sampling set minimizes the spectral radius ... |

19 | Incomplete sampling series and the recovery of missing samples from oversampled bandlimited signals - Ferreira - 1992 |

17 | Applications of reproducing kernel Hilbert spacesâ€”Bandlimited signal models - Yao - 1967 |

16 |
Convergence of iterative nonexpansive signal reconstruction algorithms
- Tom, Quatieri, et al.
- 1981
(Show Context)
Citation Context ...E ALGORITHM A number of alternative approaches to (3) do exist. The one that we will use depends upon well-known results in fixedpoint theory. This technique is often used in signal restoration [39], =-=[53]-=- and does not assume the linearity of the operators involved. In fact, other possibly nonlinear nonexpansive constraints can be incorporated in the algorithm without disturbing convergence. Denote the... |

14 |
The discrete Gerchberg algorithm
- Jones
- 1986
(Show Context)
Citation Context ...bed as an alternating projection algorithm of the type found in [4]. Its convergence will be established under no restriction on the distribution of the missing samples, generalizing results found in =-=[5]-=- and providing a theoretical background for the observations reported in [6]. It is often convenient to have upper and lower bounds for the convergence rate of an iterative algorithm. We will give the... |

14 |
Recovery of bandlimited signals from unequally spaced samples
- Wiley
(Show Context)
Citation Context ...r of a priori constraints to be included in the restoration procedure. See, for example, [23]-[26]. An iterative algorithm for restoring continuously sampled band-limited L2 functions was proposed in =-=[27]-=-. A continuously sampled signal is obtained by setting to zero a signal except for a neighborhood of a countably infinite set of sampling instants. This algorithm is related to the PapoulisGerchberg a... |

13 | The Feasible Solution in Signal Restoration - Trusell, Civanlar - 1984 |

12 |
Matrix Iterative Analysis. Englewood Cliffs
- Varga
- 1962
(Show Context)
Citation Context ...s spectral radius. Identity (1) can be established in different ways. One that leads to useful conclusions in the kind of problems that we will consider is the following (for an alternative proof see =-=[60]-=-). If we seek he stationary points of the continuous function (x) -- IlMxll 2, subject to I[xll = we are lead to the equation MHMx = Ax (2) where ,X appears as a Lagrange multiplier. This shows that t... |

12 | Restoring lost samples from an oversampled bandlimited signal - Marks - 1983 |

11 |
Iterative algorithms in irregular sampling a first comparison of methods
- Cenker, Feichtinger, et al.
(Show Context)
Citation Context ...ound in these works and in a few others which are more closely related to the subject that we are about to study. For a numerical comparison of the several existing extrapolation algorithms see [20], =-=[21]-=-. A. Band-Limited L2 and 2 Signals Many of the mappings often found in iterative constrained restoration are projections, a fact that was used [4], [11] to obtain an alternative framework for a class ... |

11 | Detection of hidden periodicities by adaptive extrapolation - Papoulis, Chamzas - 1979 |

10 | The landweber iteration and projection onto convex sets - Trussell, Civanlar - 1985 |

9 |
Unified Hilbert space approach to iterative least-squares linear signal restoration
- Sanz, Huang
- 1983
(Show Context)
Citation Context ...d integral equations. A number of two-step reconstruction algorithms can be understood in terms of this framework. A noniterative algorithm based on a distinct principle is studied in [35]. Reference =-=[36]-=- deals with iterative least-squares solutions of the linear signal restoration problem, and presents an approach to the solution of this problem based on Bialy's iteration, a general procedure to obta... |

9 | Fitting a bandlimited signal to given points - Levi - 1965 |

9 |
An iterative method to compensate for the interpolation distortion
- Marvasti
- 1989
(Show Context)
Citation Context ...these algorithms, including a class of finite-dimensional problems, was studied in [531. There are a few papers that report the performance of iterative techniques as interpolators or estimators [6], =-=[54]-=-, but the analysis normally performed does not take full advantage of the finite dimensionality of the problem. This is an important fact since it allows some of the difficulties that arise when sampl... |

9 | On the Use of Singular Value Decomposition and Decimation in Discrete-Time Band-Limited Signal Extrapolation - Sullivan - 1984 |

8 |
An extrapolation procedure for band-limited signals
- Cadzow
- 1979
(Show Context)
Citation Context ...ral existing extrapolation algorithms see [20], [21]. A. Band-Limited L2 and 2 Signals Many of the mappings often found in iterative constrained restoration are projections, a fact that was used [4], =-=[11]-=- to obtain an alternative framework for a class of restoration problems which admit an interesting geometrical interpretation [4]. In the notation of [4], x satisfies x = Pb:c, and the observed signal... |

8 |
On the properties of some systems that distort signals
- Sandberg
- 1963
(Show Context)
Citation Context ... zero a signal except for a neighborhood of a countably infinite set of sampling instants. This algorithm is related to the PapoulisGerchberg algorithm and to a general iteration proposed by Sandberg =-=[28]-=-, as discussed in [29]-[311. It was shown recently 2598 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 42, NO. 10, OCTOBER 1994 [32] that the same procedure can be used in the case of instantaneous samp... |

8 |
Optimal estimation for band-limited signals including time domain consideration
- Kolba, Parks
- 1983
(Show Context)
Citation Context ...ted functions 150]. It is possible to estimate band-limited discrete signals from a finite number of samples, taking into account the length of the interval over which a good extrapolation is desired =-=[51]-=-. Although the result does not have minimum norm, it is optimum for signals concentrated on that interval. The estimation of two-dimensional signals from a finite number of samples was also studied. I... |

8 | Some aspects of band-limited signal extrapolation: Models, discrete approximations, and noise - Sanz, Huang - 1983 |

7 | Fast Iterative and Non-Iterative Reconstruction of Band-Limited Functions from Irregular Sampling - Cenker, Feichtinger, et al. - 1991 |

7 | Discrete and continuous band-limited signal extrapolation - Sanz, Huang - 1983 |

6 | An approach to bandlimited signal extrapolation: The extrapolation matrix - Sabri, Steenaart - 1978 |

6 | Vries, "Discrete prolate spheroidal wave functions and interpolation - Delsarte, Janssen, et al. - 1985 |

5 | An Iterative Restoration Technique - Singh, Tandon, et al. - 1986 |

5 |
Reconstruction of speech signals with lost samples
- Marvasti, Clarkson, et al.
- 1992
(Show Context)
Citation Context ...ory of first-kind integral equations. A number of two-step reconstruction algorithms can be understood in terms of this framework. A noniterative algorithm based on a distinct principle is studied in =-=[35]-=-. Reference [36] deals with iterative least-squares solutions of the linear signal restoration problem, and presents an approach to the solution of this problem based on Bialy's iteration, a general p... |

5 | Estimation of sinusoids from incomplete time series - Naidu, Paramasivaiah - 1984 |

5 | A unified approach to noniterative linear signal restoration - Sanz, Huang - 1984 |

4 | On the extrapolation of band-limited functions with energy constraints - Xu, Chamzas - 1983 |

4 |
Iterative time-limited signal reconstruction
- Sanz, Huang
- 1983
(Show Context)
Citation Context ...in this way. It is also known [37] that the Papoulis-Gerchberg algorithm can be obtained from Landweber's iteration, an iterative method for the solution of first-kind Fredholm integral equations. In =-=[38]-=- it is shown that time-limited restoration of shiftinvariant blurred 2 signals can be done iterating an operator equation which defines a contractive mapping. The results are valid for multi-dimension... |

4 | An improved version of PapoulisGerchberg algorithm on band-limited extrapolation - Chamzas, Xu - 1984 |

3 |
On the extrapolation of optical image data
- Viano
- 1976
(Show Context)
Citation Context ... [2], [3], brought a new interest to the problem, and, since then, several works dealing with different aspects of the extrapolation problem have appeared. We refer the interested reader to [1], [7], =-=[9]-=--[19] among many others. These works address topics that include implementation aspects, numerical stability, convergence acceleration, noniterative extrapolation, the sampled analog of the extrapolat... |

3 | Error analysis of a class of constrained iterative restoration algorithms - Yeh, Chin - 1985 |

3 |
Numerical comparison of several algorithms for band-limited signal extrapolation
- Huang, Sanz, et al.
- 1984
(Show Context)
Citation Context ...ults found in these works and in a few others which are more closely related to the subject that we are about to study. For a numerical comparison of the several existing extrapolation algorithms see =-=[20]-=-, [21]. A. Band-Limited L2 and 2 Signals Many of the mappings often found in iterative constrained restoration are projections, a fact that was used [4], [11] to obtain an alternative framework for a ... |

3 | Convergence criteria for iterative restoration methods - Trussel - 1983 |