341
Views
3
CrossRef citations to date
0
Altmetric
Articles

Spectral method for ill-posed problems based on the balancing principle

, &
Pages 292-306 | Received 08 May 2013, Accepted 10 Feb 2014, Published online: 20 Mar 2014

Abstract

In this article, we consider a class of ill-posed problems in two-dimensional setting which can be formulated as the problem of solving ill-posed operator equation. The spectral method is a simple and effective regularization method for such a class of ill-posed problems. For the spectral method, we obtain the error estimates with optimal order under an adaptive a posteriori parameter choice rule called balancing principle. Two ill-posed examples including an inverse diffraction problem and a deblurring problem are provided.

1 Introduction

Many PDE-based ill-posed problems in mathematical physics can be formulated as the problem for solving an operator equation, e.g. the inverse diffraction problem [Citation1Citation4] and the deblurring problem.[Citation5, Citation6] Therefore, theory analysis, such as error estimate for the spectral method, is possible for this kind of ill-posed problems. In this paper, we discuss this kind of ill-posed problems in two-dimensional space, which involves the numerical computation of pseudodifferential operator.

Let1.1 h^(ξ,η)=12πR2h(x,y)e-i(ξx+ηy)dxdy1.1 be the Fourier transform of the function h(x,y)L2R2. The Sobolev function space HpR2 is defined by1.2 HpR2=h(x,y)|hL2R2,hp:=R21+ξ2+η2p|h^ξ,η|2dξdη12<,1.2 where h^(ξ,η) is the Fourier transform of function h(x,y) and ·:=·0 denotes the norm in L2R2.

Consider the numerical computation of the pseudodifferential operator with an unbounded symbol a(ξ,η,s) given by1.3 f(x,y,s)=12πR2a(ξ,η,s)g^(ξ,η)ei(ξx+ηy)dξdη,1.3 where s(0,L] is a fixed constant and g^(ξ,η) is the Fourier transform of the exact data g(x,y)L2R2; the data g(x,y) is only given approximately by gδ(x,y)L2R2 satisfying1.4 g(·,·)-gδ(·,·)δ,1.4 where δ is the noise level and is assumed to be known.

We assume that the symbol a(ξ,η,s) satisfies the following growth condition:1.5 C1expsϕξ2+η2|aξ,η,s|C2expsϕξ2+η2,ξ2+η2c1,s(0,L],|a(ξ,η,s)|z(s),ξ2+η2<c1,1.5 where the function ϕ(x) is strictly increasing with limxϕ(x)=, the function z(s)C[0,L] with z(s)>0 and C1,C2,c1>0 are constants.

If a(ξ,η,s) satisfies |a(ξ,η,s)|=O(exp(c(ξ2+η2)b)) as ξ2+η2 with c,b>0, then we call the problem of numerical computation of the above pseudodifferential operators as a severely ill-posed problem. Likewise, if a(ξ,η,s) satisfies |a(ξ,η,s)|=Oξ2+η2b as ξ2+η2 with b>0, then we call the problem of numerical computation of above pseudodifferential operators as a mildly ill-posed problem. In this paper, we only consider the severely ill-posed problem since the mildly ill-posed problems can be solved more easily. The problem (Equation1.3) in one-dimensional space has been discussed in [Citation7] by the a posteriori Fourier method based on Morozov’s discrepancy principle. But, many ill-posed problems with important application backgrounds such as inverse diffraction and deblurring problems are set in two-dimensional case. This motivates us to consider the two-dimensional inverse problems. Although the authors in [Citation7] analysed the Fourier method with a posteriori parameter choice rule by using a direct method, here we stress that we give a rule for selecting regularization parameter based on the theory of regularization framework. The analysis involved is more general.

A priori information about unknown solution has been proved to be essential in the analysis of ill-posed problems in mathematical physics. Otherwise, without the a priori information, the convergence rate of the constructed regularization method will be arbitrarily slow.[Citation8] For analysis, we assume that there holds the a-priori bound for the unknown solution1.6 f(·,·,L)pE,p0.1.6 Let fαδ(·,·,s) denote a regularization solution. If 0s<L, we call the fαδ(·,·,s) as interior inversion. If s=L, we call the fαδ(·,·,s) as boundary inversion.

Obviously, the ill-posedness of problem (Equation1.3) is caused by components of high frequency. If c10, we only need to regularize the solution in the case of ξ2+η2c1, because for the case of ξ2+η2c1 the problem (Equation1.3) is well posed which can be observed by the following discussion:

Let the sets W(ξ,η)={(ξ,η)|ξ2+η2c1} and I(ξ,η)={(ξ,η)|ξ2+η2c1}, and we have L2R2=L2(W)L2(I).

Consider the differencef^(ξ,η,s)-f^δ(ξ,η,s)2L2(W)=W|a(ξ,η,s)|2|g^(ξ,η)-g^δ(ξ,η)|2dξdηz2(s)δ2Cδ2,where f^δ(ξ,η,s) denotes the solution with noisy data gδ.

The same technique has been used in.[Citation9, Citation10] Throughout this paper, we mainly deal with the case of ξ2+η2c1.

By the balance principle,[Citation11] we can derive a Hölder-type error estimate like fαδ(·,·,s)-f(·,·,s)=O(Es/Lδ1-sL),δ0 under the a priori bound (Equation1.6) with p=0. However, when s=L, in which we should be more interested, we cannot get any convergence. To overcome the difficulty, usually a stronger a priori bound on the unknown solution is added, i.e. we should require p>0 for (Equation1.6) to obtain the convergence rate at s=L. In the following, we will construct the interior and boundary inversions by a spectral regularization method.

The outline of this paper is as follows. In Section 2, we present the spectral method. In Section 3, the spectral method is applied to solve problem (Equation1.3); some applications are given in Section 4. In Section 5, some numerical results are presented. Finally, Section 6 concludes the paper.

2 The spectral method [Citation12, Citation13]

2.1 Preliminaries

Ill-posed operator equations arise in several contexts and various aspects have been treated in the literature.[Citation8, Citation14, Citation15] We cannot give here an exhaustive survey. In this paper, we are interested in solving the solution xH1 of linear ill-posed problems by spectral cut-off method. Consider2.1 Ax=y,2.1 where A:H1H2 is a linear injective, closed operator between infinite-dimensional Hilbert spaces H1 and H2 with non-closed range R(A). We suppose that yδH2 are the noisy data with2.2 y-yδδ2.2 and known noise level δ.

If we consider the ill-posed problem Kx=y where only noisy data yδ are available, K is a compact operator with singular system {σn,vn,un}n=1 since the ill-posedness is associated with the small singular values, an obvious idea is to truncate or damp the smaller singular values. That is the well-known truncated singular value decomposition method. These methods are simple and effective. In practical computation, we can implement the spectral cut-off method by mollification method discovered by Hào [Citation16].

For problem (Equation2.1), most regularization operators can be written in the form,2.3 Rα:=gα(AA)A2.3 with some function gα satisfyinglimα0gα(λ)=1λ,where the operator function gα(AA) is well defined via the spectral representation gα(AA)=0ag(λ)dEλ. Here, AA=0aλdEλ, {Eλ} denotes the spectral family of the operator AA and a is a constant satisfying AAa with a= if AA is unbounded.

Then, for the regularization solution with unperturbed data, we have xα:=Rαy and x-xα=rα(AA)x with rα(λ)=1-λgα(λ). For example, for spectral cut-off method,2.4 gα(λ)=1λ,λα,0,λ<α.2.4 In general, the exact solution xX is required to satisfy a so-called source condition, otherwise the convergence of the regularization method approximating the problem can be arbitrarily slow. For ill-posed problems, the source condition is chosen as2.5 x=φ(AA)1/2ω,ωE,2.5 i.e. x belongs to the source set2.6 Mφ,E=φ(AA)1/2ω,ωXandωE,2.6 where φ(λ) satisfies some properties:

Assumption 1

limλ0φ(λ)=0andφ(λ)is strict monotonically increasing, ρ(λ)=λφ-1(λ)is convex.

For the stable approximate solution of problem (Equation2.1) some regularization technique has to be applied, which provides regularized approximations xαδ=Rαδyδ with property xαδx as δ0.

Any operator R:H2H1 can be considered as a special method for solving problem (Equation2.1). the approximate solution to (Equation2.1) is then given by Ryδ. Consider the worst case error Δ(δ,R) for identifying the solution x of problem (Equation2.1) from noisy data yδ under the assumptions y-yδδ and x belongs to a source set Mφ,E which is defined by2.7 Δ(δ,R)=supRyδ-x|xMφ,E,yδH2,y-yδδ.2.7 This worst case error characterizes the maximal error of the method R if the solution x of problem (Equation2.1) varies in the set Mφ,E. An optimal method Ropt is characterized by Δ(δ,Ropt)=infRΔ(δ,R). It can easily be realized that2.8 infRΔ(δ,R)ω(δ,Mφ,E),2.8 where ω(δ,Mφ,E)=sup{x|xMφ,E,Axδ}.

The following theorem and definition can be found in [Citation15].

Theorem 1.1

Let Mφ,E is given by (Equation2.6), and Assumption 1 be satisfied. If δEσ(AAφ(AA)), then2.9 ωδ,Mφ,E=Eρ-1δ2E2,2.9 where ρ is given by ρ(λ)=λφ-1(λ).

Definition 1.1

Let Assumption 1 be satisfied. Any regularization method Rαδ for problem (Equation2.1) with noisy data is called

(i)

Optimal on the set Mφ,E if xαδ-xEρ-1(δ2E2);

(ii)

Order optimal on the set Mφ,E if xαδ-xcEρ-1(δ2E2) with c1.

2.2 The balancing principle

Now, we prove some estimates for the spectral regularization method under the a posteriori parameter choice rule based on balance principle.

First we have [Citation12]:

Theorem 2.1

Let xMφ,E, Assumption 1 is satisfied, then there holds2.10 xαδ-xEφ(α)+δ/α.2.10

Now denote C(α):=Eφ(α), D(α):=δ/α. Since C(α) increases with α but D(α) decreases with α, let us rewrite (Equation2.10) as2.11 xαδ-xC(α)+D(α).2.11 By virtue of the behaviour of the functions C(α) and D(α), we can choose an α~ such that C(α~)=D(α~). Therefore,2.12 xαδ-x2C(α~).2.12 However, since the a priori bound E of the unknown solution is seldom known such that C(α) is unknown, the a priori choice of the theoretically value of α~ is impossible. Therefore it is necessary to use some a-posteriori rule of α.

For this purpose, we use the balancing principle.

Consider a discrete set of possible values of regularization parameter2.13 N=αi=q2iα0,i=1,2,,N,q>1,2.13 where α0=δ2, N sufficiently large such that αN1.

Denote the set2.14 M+N=αiN:xαiδ-xαjδ4D(αj),j=1,,i.2.14 Choose the value of regularization parameter by the rule2.15 α=α+:=maxαM+(N).2.15 Below we show that the choice of α=α+ gives the error estimate, which differs from (Equation2.12) only by a factor of 3q.

To prove this conclusion, denote the auxiliary set2.16 MN=αiN:CαjD(αi),j=1,,i.2.16 and the auxiliary quantity2.17 α:=maxαM(N).2.17 Without loss of generality, we assume thatMN,N\MN.

Theorem 2.2

Let the parameter be chosen by (Equation2.15), then there holds2.18 xα+δ-x6qC(α~)=6qEφ(α~).2.18

Proof

As the process is similar to [Citation11], we omit it.

3 Spectral regularization for problem (Equation1.3)

Now, in order to apply the spectral regularization method, we formulate problem (Equation1.3) as an operator equation in the frequency domain:A(s)f^:=a-1(ξ,η,s)f^(ξ,η,s)=g^(ξ,η),where A(s):L2R2L2R2 is a multiplication operator. Obviously A(s)=a-1(ξ,η,s), and the adjoint operator of A(s) is A(s)=a-1(ξ,η,s)¯, where the symbol Π¯ denotes the complex conjugate of Π. Therefore, AA(s)=|a-1(ξ,η,s)|2=|a(ξ,η,s)|-2. In this section, we deal with problem (Equation1.3) for two cases separately:

Case I    (the case of the interior inversion): p=0 in (Equation1.6).

Case II    (the case of the boundary inversion): p>0 in (Equation1.6).

Now we investigate the spectral method for two cases.

First, we can write the spectral regularization according to (Equation2.4):3.1 f^αδ(ξ,η,s)=a(ξ,η,s)g^δ(ξ,η),|a(ξ,η,s)|1α,0,else.3.1 In order to obtain the explicit expression of the error bound, we need to know the expression of φ(·) in (Equation2.5).

3.1 Interior inversion

In this subsection, we note that p=0 in (Equation1.6) and we want to recover the solution f^(ξ,η,s) with 0s<L. First we use the equalityf^(ξ,η,s)=a(ξ,η,s)a(ξ,η,L)f^(ξ,η,L).Now (Equation1.6) readsf^(ξ,η,L)2=a(ξ,η,L)a(ξ,η,s)f^(ξ,η,s)2E2.Denote the sets3.2 M1=f^(ξ,η,s)L2|a(ξ,η,L)a(ξ,η,s)f^(ξ,η,s)2E2.3.2 3.3 M=f^(ξ,η,s)L2|C1C22|a(ξ,η,s)|L-ssf^(ξ,η,s)2E2.3.3 In fact, by AA(s)=|a-1(ξ,η,s)|2=|a(ξ,η,s)|-2, M is equivalent to M2 given in the form of (Equation2.6)3.4 M2=f^(ξ,η,s)L2|φ(AA(s))-12f^(ξ,η,s)2E2,3.4 where3.5 φ(λ)=C24C12λL-ss.3.5 Now we only need to show that every element from M1 belongs to the set M. For any element f^(ξ,η,s)M1, we have3.6 a(ξ,η,L)a(ξ,η,s)2f^(ξ,η,s)2dξE2.3.6 By (Equation1.5), this implies3.7 C1C2exp(L-s)ϕξ2+η22f^ξ,η,s2dξE2.3.7 On the other hand, for any element f^(ξ,η,s)M, we have two-side estimatesC12C24|C1exp(L-s)ϕξ2+η2|2|f^(ξ,η,s)|2dξdηC12C24|a(ξ,η,s)|2(L-s)s|f^(ξ,η,s)|2dξdηC12C24|C2exp(L-s)ϕξ2+η2|2|f^(ξ,η,s)|2dξdηE2.Thus, we have shown that every element from M1 belongs to the set M=M2. Likewise, we can also show that every element from M0 belongs to the set M1 by constructing another set M0 similar to M. We have another φ(·) which differs from (Equation3.5) only by a constant factor. Now we summarize what we have:

Conclusion 1

If the a-priori bound (Equation1.6) with p=0 holds, for problem (Equation1.3), the function φ(·) in the source set (Equation3.5) has the form of φ(λ)=γλL-ss with γ is a constant which only depends on C1 and C2. Thus, the function ρ(λ)=λφ-1(λ)=c2λLL-s and ρ-1(λ)=c3λL-sL. Therefore for problem (Equation1.3) for solving f^(ξ,η,s), the optimal convergence order for error estimates is given by3.8 E(δ/E)L-sL=EsLδL-sL.3.8

3.2 Boundary inversion

In this subsection, we note that p>0 in (Equation1.6) and we want to recover the solution f^(ξ,η,L). Now (Equation1.6) with p>0 readsf^(ξ,η,L)p2=1+ξ2+η2p/2f^(ξ,η,L)2E2.Denote the sets3.9 M3=f^(ξ,η,L)L2|1+ξ2+η2p/2f^(ξ,η,L)2E2.M4=f^(ξ,η,L)L2|ϕ-1γ02LlnAA(L)-1pf^(ξ,η,L)2E2,3.9 where the ϕ-1(·) denotes the inverse function of ϕ(·) and γ0 is a positive constant depending on C1,C2. Similar to Conclusion 1, we can prove that every element from M3 belongs to the set M4. Thus, we can obtain the expression of φ(λ):3.10 φ(λ)=ϕ-1γ02Lln([λ]-1)-p.3.10 By an elementary calculation, it yields3.11 ρ-1(λ)=ϕ-1c~lnλ-1-p,3.11 where c~ is a constant depends on γ0,L.

Conclusion 2

If the a-priori bound (Equation1.6) with p>0 holds, the optimal convergence order for solving f^(ξ,η,L) is given by3.12 ϕ-1c~lnE2/δ2-p.3.12

4 Applications

Now, we give some applications of Theorem 2.3. Throughout this section, the c~,c are the positive constants which are not dependent on δ and E.

Example 4.1

The reconstruction problem of aperture in the plane from their diffraction patterns arises in acoustics and optics. Let us consider the following inverse diffraction problem. Let D be a bounded aperture in an infinite perfectly soft screen which is located in the plane z=0 in R3. A harmonic plane wave with wave-number k propagates along with the positive z direction. It hits the screen and escapes through the aperture D. The measured data at receiving screen z=L>0 are given. The problem is to reconstruct the shape (domain) of the aperture D. By Kirchhoff approximation, the mathematical modelling can be formulated as follows.

Let u(x,y,z) be the solution of the following problem:4.1 uxx+uyy+uzz+k2u=0,(x,y)R2,z>04.1 4.2 u(x,y,L)=g(x,y),(x,y)R2,4.2 4.3 limrrur-iku=0,wherer=x2+y2+z2.4.3

Here we wish to find the u(x,y,0)=χD(x,y) from the measured data gδ(x,y)L2, where χD(x,y) is the characteristic function of the domain D.

As usual, we assume that there exists a priori bound:4.4 u(·,·,0)pE,4.4 where ·p denotes the norm of Sobolev space Hp(R2) with p0 and E is a positive constant.

We have the solution in the frequency domain:4.5 χD^(ξ,η)=eLξ2+η2-k2g^(ξ,η).4.5 Therefore, the symbol of pseudodifferential operator is given by4.6 a(ξ,η,L)=eLξ2+η2-k2.4.6 Now let ϕ(ω)=ω-k2 and its inverse function ϕ-1(ω)=ω2+k2, f(x,y,L)=u(x,y,0)=χD(x,y); by (Equation3.11), we have4.7 φ(α~)=ϕ-1γ02Llnα~-1-p=OEln1α~-2p,α~0.4.7 Therefore, by Theorem 2.3, we have the following error estimate:4.8 fα+δ(·,·,L)-u(·,·,0)cEln1α~-p.4.8 Now we need to estimate α~. At α=α~, we have C(α~)=D(α~). Therefore,cEln1α~-p=δ/α~,i.e.α~=δ2(cE)2ln2p1α~.Since lnmx<xυ holds for any x>0,υ>0,m>0, we haveα~=δ2(cE)2ln2p1α~δ2(cE)21α~,i.e.4.9 α~δcE.4.9 As a result, we have the error estimate4.10 fα+δ(·,·,L)-u(·,·,0)c~ElnEδ-p.4.10 The convergence order is optimal.

Example 4.2

Consider the linear diffusion final value problem in L2R2,4.11 ut(x,y,t)=Δu(x,y,t),t>0,u(x,y,T)=g(x,y),4.11 where Δ is the Laplacian operator, usually g(x,y)L2R2 are the data with approximation gδ(x,y)L2R2. As usual, we assume that there exists a priori bound:4.12 u(·,·,0)pE,4.12 where ·p denotes the norm of Sobolev space HpR2 with p0 and E is a positive constant.

This problem can be associated with a deblurring problem. Consider a space invariant point spread functions p(x,y) and formulate the deblurring problem as4.13 R2p(x-μ,y-ν)f(μ,ν)dμdν=g(x,y),4.13 where f(x,y) is the desired unblurred image and g(x,y) is the blurred image that would have been recorded in the absence of noise. In general, the image f(x,y) is viewed as originally defined on a bounded rectangle-like domain; we could extend f(x,y) periodically to all of R2.

We can easily get the unique Fourier space solution for problem (Equation4.11)4.14 u^(ξ,η,t)=exp(T-t)ξ2+η2g^(ξ,η),t>0,4.14 If we take p^(ξ,η)=exp(T(ξ2+η2)),T>0, we may identity the blurred image g(x,y) with the temperature distribution at time t=T in an infinite two-dimensional medium whose initial temperature distribution is given by the unblurred image u(x,y,0):=f(x,y). This involves solving the diffusion equation backwards in time. For a fixed 0<t<T, we call the solution u(x,y,t) a partial restoration. For the case t=0, we call the solution u(x,y,0) a full restoration.

Therefore, the symbol of pseudodifferential operator is given by4.15 a(ξ,η,t)=exp(T-t)ξ2+η2.4.15 Now let s=T-t, L=T, ϕ(ω)=ω and its inverse function ϕ-1(ω)=ω, f(x,y,L)=u(x,y,0).

Boundary inversion. By (Equation3.11), we have4.16 φ(α~)=ϕ-1γ02Llnα~-1-p=OEln1α~-p,α~0.4.16 Therefore, similar to Example 4.1, by Theorem 2.3, we have the following error estimate:4.17 fα+δ(·,·,L)-u(·,·,0)cElnEδ-p/2.4.17 Interior inversion. By (Equation3.5), we have4.18 φ(α~)=OEα~t2(T-t),α~0.4.18 Therefore, by Theorem 2.3, we have the following error estimate:4.19 fα+δ(·,·,s)-u(·,·,t)cEα~t2(T-t).4.19 Now we need to estimate α~. At α=α~, we have C(α~)=D(α~). Therefore,cEα~t2(T-t)=δ/α~,i.e.α~=δcE2(T-t)T.As a result, we have the error estimate4.20 fα+δ(·,·,L)-u(·,·,0)c~δt/TE1-t/T.4.20 The convergence order in (Equation4.17) and (Equation4.20) are optimal.

5 Numerical results

In this section, we only consider the numerical results of Example 4.1. The numerical results of Example 4.2 can be found in [Citation6], we do not repeat them here. Although the examples are described in an unbounded domain in the plane (x,y), we are interested in the domain (x,y)[0,1]×[0,1]. This is reasonable because the problem can be solved by periodic extension in the (x,y) plane.

Define the discrete L2 norm of u(x,y) by5.1 u=n-2j,k=1nu(xj,yk)212.5.1 where n is the total number of sampled points. In this section, we take n=100.

In order to measure the accuracy of numerical results, we define the discrete L2-norm error for the exact solution u between the approximate solution ua as follows:5.2 E=n-2j,k=1n|u(xj,yk)-ua(xj,yk)|212.5.2 The noise was added to g(x,y) by setting δ(xj,tk)=σrjkmax{g(xj,yk)}), where σ denotes the noise level and rjk is a random number drawn from a uniform distribution in the range [0,1]. In numerical experiment, we use (Equation5.2) to compute δ=g-gδ.

Figure 1. The original D (a) and the data g (b).

Figure 1. The original D (a) and the data g (b).

Figure 2. The reconstructed D by spectral method.

Figure 2. The reconstructed D by spectral method.

We shall illustrate the reconstruction method (Equation3.1) with different numerical examples although we have used some ‘inverse crimes’.[Citation17] The Fourier formula (Equation3.1) is based on FFT algorithm.

Example 1

We choose the aperture D in the shape of the letter ‘L’ for our test. The characteristic function, as shown in Figure (a) on the domain D, is to be constructed. And Figure (b) displays data g which is a very blurred picture of the original aperture.

In this example, 1% relative error is added to the data g(x,y) at the sampled points in the simulation and we can get δ=0.02. In the computation, we used q=1.2, α+=4.710-5. Figure corresponds to the results obtained by the spectral method, where the distance from the receiving plane to the screen is L=0.2 with the wave-number k=5.

Figure 3. The original D (a) and the data g (b).

Figure 3. The original D (a) and the data g (b).

Figure 4. The reconstructed D by spectral method.

Figure 4. The reconstructed D by spectral method.

Example 2

The aperture D consists of three independent squares. The characteristic function to be constructed and the data g are displayed in Figure .

In this example, 1% relative error is added to the data g(x,y) at the sampled points in the simulation as the Example 5.1. And we can get δ=0.02. In the computation, we used q=1.2, α+=1.110-4.

Figure correspond to the results by the spectral method, where the distance from receiving plane to the screen is L=0.1 with the wave-number k=5.

In the numerical tests, we find that the reconstruction gets more severe as the receiving plane is moved away from the aperture and as the wave-number k is decreased. Because if k is decreased and d is increased, the degree of ill-posedness of the problem increases.

6 Concluding remark

Numerical solutions for many ill-posed problems can be recast as the numerical computation of a class of pseudodifferential operators. In this paper, for the numerical computation of a class of pseudodifferential operators, we investigate the spectral method. Under the balancing principle, we derived the error bounds for the numerical computation of a class of pseudodifferential operators. The obtained error bounds have the optimal convergence order. Two numerical examples are provided and the numerical experiment shows that the method works well.

Acknowledgments

The reviewers are thanked for their very careful reading and for pointing out several mistakes as well as for their useful comments and suggestions.

Notes

1 The work described in this paper was partially supported by a grant from the National Natural Science Foundation of China [grant number 11001223]; the Research Fund for the Doctoral Program of Higher Education of China [grant number 20106203120001]; the Key (Keygrant) Project of Chinese Ministry of Education [grant number 212179]; and the Doctoral Foundation of Northwest Normal University, China [grant number 5002–577].

References

  • Magnanini R, Papi G. An inverse problem for the Helmholtz equation. Inverse Probl. 1985;1:357–370.
  • Sondhi M. Reconstruction of objects from their sound-diffraction patterns. J. Acoust. Soc. Am. 1969;46:1158–1164.
  • Bertero M, De CM. Stability problems in inverse diffraction. IEEE Trans. Antenn. Propag. 1981;AP-29:368–372.
  • Santosa F. A level-set approach for inverse problems involving obstacles. Control Optim. Cal. Var. 1996;1:17–33.
  • Carasso AS. Overcoming Hölder continuity in ill-posed continuation problems. SIAM J. Numer. Anal. 1994;31:1535–1557.
  • Xiong XT, Wang JX, Li M. An optimal method for fractional heat conduction problem backward in time. Appl. Anal. 2012;91:823–840.
  • Fu CL, Zhang YX, Cheng H, Ma YJ. The a posteriori Fourier method for solving ill-posed problems. Inverse Probl. 2012;28:095002. 26 p.
  • Engl HW, Hanke M, Neubauer A. Regularization of inverse problems. Dordrecht: Kluwer Academic Publisher; 1996.
  • Xiong XT, Fu CL. Two approximate methods of a Cauchy problem for the Helmholtz equation. Comput. Appl. Math. 2007;26:285–307.
  • Reginska T, Tautenhahn U. Conditional stability estimates and regularization with applications to Cauchy problems for the Helmholtz equation. Numer. Funct. Anal. Optim. 2009;30:1065–1097.
  • Pereverzev S, Shock E. On the adaptive selection of the parameter in regularization of ill-posed problems. SIAM J. Numer. Anal. 2006;75:2060–2076.
  • Xiong XT, Zhao XC, Wang JX. Spectral Galerkin method and its application to a Cauchy problem of Helmholtz equation. Numer. Algor. 2013;63:691–711.
  • Xiong XT, Fu CL, Qian Z. On three spectral regularization methods for a backward heat conduction problem. J. Korean Math. Soc. 2007;44:1281–1290.
  • Nair MT. Linear operator equations: approximation and regularization. Singapore: World Scientific; 2009.
  • Tautenhahn U. Optimality for linear ill-posed problems under general source conditions. Numer. Funct. Anal. Optim. 1998;19:377–398.
  • Hào DN. Methods for inverse heat conduction problems. Frankfurt am Main, New York (NY): Peter Lang; 1998.
  • Kaipio JP, Somersalo E. Statistical and computational inverse problems. New York (NY): Springer, Applied Mathematical Sciences; 2005.

Reprints and Corporate Permissions

Please note: Selecting permissions does not provide access to the full text of the article, please see our help page How do I view content?

To request a reprint or corporate permissions for this article, please click on the relevant link below:

Academic Permissions

Please note: Selecting permissions does not provide access to the full text of the article, please see our help page How do I view content?

Obtain permissions instantly via Rightslink by clicking on the button below:

If you are unable to obtain permissions via Rightslink, please complete and submit this Permissions form. For more information, please visit our Permissions help page.