In this paper, a new total generalized variational (TGV) model for restoring images with Cauchy noise is proposed, which contains a non-convex fidelity term and a TGV regularization term. In order to obtain a strictly convex model, we add an appropriate proximal term to the non-convex fidelity term. We prove that the solution of the proposed model exists and is unique. Due to the convexity of the proposed model and in order to get a convergent algorithm, we employ an alternating minimization algorithm to solve the proposed model. Finally, we demonstrate the performance of our scheme by numerical examples. Numerical results demonstrate that the proposed algorithm significantly outperforms some previous methods for Cauchy noise removal.
Citation: Lufeng Bai. A new approach for Cauchy noise removal[J]. AIMS Mathematics, 2021, 6(9): 10296-10312. doi: 10.3934/math.2021596
[1] | Taha Gökmen Ülger, Muhittin Tayfur, Funda Pınar Çakıroğlu, Çiğdem Özcan . The role of duodenal jejunal bypass liner in obesity treatment. AIMS Medical Science, 2021, 8(3): 224-236. doi: 10.3934/medsci.2021019 |
[2] | Sanaa El–Jamal, Houda Elfane, Imane Barakat, Khadija Sahel, Mohamed Mziwira, Aziz Fassouane, Rekia Belahsen . Association of socio-demographic and anthropometric characteristics with the management and glycemic control in type 1 diabetic children from the province of El Jadida (Morocco). AIMS Medical Science, 2021, 8(2): 87-104. doi: 10.3934/medsci.2021010 |
[3] | Hidetaka Hamasaki . Dog ownership: is it beneficial for physical activity, cardiovascular disease, and diabetes?. AIMS Medical Science, 2020, 7(4): 311-327. doi: 10.3934/medsci.2020021 |
[4] | Snigdha Misra, Yang Wai Yew, Tan Seok Shin . Maternal dietary patterns, diet quality and micronutrient status in gestational diabetes mellitus across different economies: A review. AIMS Medical Science, 2019, 6(1): 76-114. doi: 10.3934/medsci.2019.1.76 |
[5] | Valeria Hirschler . Skin and obesity in childhood: an update. AIMS Medical Science, 2021, 8(4): 311-323. doi: 10.3934/medsci.2021026 |
[6] | Burcin Meryem Atak Tel, Ramiz Tel, Tuba Duman, Satilmis Bilgin, Hamza Kaya, Halil Bardak, Gulali Aktas . Evaluation of the relationship between the prealbumin/fibrinogen ratio and diabetic nephropathy in patients with type 2 diabetes mellitus. AIMS Medical Science, 2024, 11(2): 90-98. doi: 10.3934/medsci.2024008 |
[7] | Neimat Mahmoud Abd-Alrahman Ali Dinar, Ghassan Abd-Al lateef Mohammad Al Sammouri, Mohamed Abdalla Eltahir, Aida Ahmed Fadlala Ahmed, Hasen Jamaan Ahmed Alghamdi, Abdulrahman Ali Alghamdi, Waled Amen Mohammed Ahmed . Effect of diabetes educational program on self-care and diabetes control among type 2 diabetic patients in Al-Baha–Saudi Arabia. AIMS Medical Science, 2019, 6(3): 239-249. doi: 10.3934/medsci.2019.3.239 |
[8] | Hidetaka Hamasaki . Functional Foods for Type 2 Diabetes. AIMS Medical Science, 2016, 3(3): 278-297. doi: 10.3934/medsci.2016.3.278 |
[9] | Kevin Rebecchi . Beyond “autism spectrum disorder”: toward a redefinition of the conceptual foundations of autism. AIMS Medical Science, 2025, 12(2): 193-209. doi: 10.3934/medsci.2025012 |
[10] | Sylvia Kirchengast, Beda Hartmann . Maternal prepregnancy nutritional status influences newborn size and mode of delivery. AIMS Medical Science, 2018, 5(1): 53-66. doi: 10.3934/medsci.2018.1.53 |
In this paper, a new total generalized variational (TGV) model for restoring images with Cauchy noise is proposed, which contains a non-convex fidelity term and a TGV regularization term. In order to obtain a strictly convex model, we add an appropriate proximal term to the non-convex fidelity term. We prove that the solution of the proposed model exists and is unique. Due to the convexity of the proposed model and in order to get a convergent algorithm, we employ an alternating minimization algorithm to solve the proposed model. Finally, we demonstrate the performance of our scheme by numerical examples. Numerical results demonstrate that the proposed algorithm significantly outperforms some previous methods for Cauchy noise removal.
The stability problem for the functional equation was conjectured by Ulam [1] in 1940. In the following year, Hyers [2] presented a partial answer for the case of the additive mapping in this problem: If f satisfies |f(x+y)−f(x)−f(y)|≤ε for some fixed ε>0, then there is an additive mapping g satisfying g(x+y)=g(x)+g(y) and |f(x)−g(x)|≤ε, which is called the Hyers-Ulam stability.
Baker et al. [3] announced, in 1979, the new concept for the superstability as follows: If f satisfies |f(x+y)−f(x)f(y)|≤ε for some fixed ε>0, then either f is bounded or f satisfies the exponential functional equation f(x+y)=f(x)f(y).
Baker [4] showed the superstability of the cosine (also called d'Alembert) functional equation
f(x+y)+f(x−y)=2f(x)f(y). | (C) |
The cosine (d'Alembert) functional equation (C) was generalized to the following:
f(x+y)+f(x−y)=2f(x)g(y), | (W) |
f(x+y)+f(x−y)=2g(x)f(y), | (Kgf) |
f(x+y)−f(x−y)=2f(x)f(y), | (T) |
f(x+y)−f(x−y)=2g(x)f(y), | (Tgf) |
f(x+y)−f(x−y)=2g(x)g(y), | (Tgg) |
f(x+y)−f(x−y)=2g(x)h(y), | (Tgh) |
in which (W) is called the Wilson equation, (Kgf) is called the Kim's equation and remaining equations are raised in Kim's papers [5,6,7].
The superstability of the trigonometric (cosine (C), Wilson (W), Kim ((Kgf), (T), (Tgf) and (Tgh)) functional equations were founded in Badora [8], Badora and Ger [9], Kannappan and Kim [10], Kim and Dragomir [11], Kim [5,6,7] and in papers [13,14,15].
In 1983, Cholewa [16] proved the superstability of the sine functional equation
f(x+y2)2−f(x−y2)2=f(x)f(y) | (S) |
under the stability inequality bounded by constant. This was improved to the condition bounded by a function in Badora and Ger [9]. Their results were also further improved later by Kim [17,18], who obtained the superstability under the assumption that the stability inequality is bounded by a constant or a function for the generalized sine functional equations:
f(x+y2)2−f(x−y2)2=f(x)g(y), | (Sfg) |
f(x+y2)2−f(x−y2)2=g(x)f(y), | (Sgf) |
f(x+y2)2−f(x−y2)2=g(x)g(y), | (Sgg) |
f(x+y2)2−f(x−y2)2=g(x)h(y). | (Sgh) |
In 2009, Eshaghi Gordji and Parviz [19] introduced the quadratic-radical functional equation
f(√x2+y2)=f(x)+f(y). | (R) |
related to the additive mapping and proved its stability.
Recently, Almahalebi et al. [20], Kim [21,22] obtained the superstability of p-radical functional equations in relation with Wilson (W), Kim ((Kgf), (Tgf) and (Tgh)). In the concept of the p-radical, the sine functional equation (S) and (S)-type's equations (Sfg), (Sgf), (Sgg), (Sgh) are expressed as follows:
f(p√xp+yp2)2−f(p√xp−yp2)2=f(x)f(y), | (Sr) |
f(p√xp+yp2)2−f(p√xp−yp2)2=f(x)g(y), | (Srfg) |
f(p√xp+yp2)2−f(p√xp−yp2)2=g(x)f(y), | (Srgf) |
f(p√x2+y22)2−f(p√x2−y22)2=g(x)g(y), | (Srgg) |
f(p√xp+yp2)2−f(p√xp−yp2)2=g(x)h(y). | (Srgh) |
In the above, letting f(x)=F(xp), then F satisfies (S)-type equations. Hence, in this paper, they will be reasonably called the p-power-radical equation. Since the function f(x)=sinxp is the solution of the equation (Sr), it will be called the p-power-radical sine functional equation.
Our aim of this paper is to find solutions and to investigate the superstability bounded by a function (Gǎvruta sense) for the p-power-radical sine functional equation (Sr) from an approximation of the p-power-radical functional equation (Srgh).
As a corollary, we obtain the superstability bounded by a constant and a function for the p-power-radical sine functional equation (Sr) from an approximation of the p-power-radical functional equations (Sr), (Srgf), (Srfg), (Srgg). Moreover, the obtained results are extended to Banach algebras.
In this paper, let R be the field of real numbers, R+=[0,∞) and C be the field of complex numbers. We assume that f,g,h are nonzero functions, ε is a nonnegative real number, φ:R→R+ is a given nonnegative function and p is a positive odd integer.
Let's recall the trigonometric formula, the p-power-radical functional equation's forms for the functional equations (cosine (d'Alembert) (C), Wilson (W), Kim (Tgf), (Tgg) and (Tgh) are the following:
f(p√xp+yp)+f(p√xp−yp)=2f(x)f(y), | (Cr) |
f(p√xp+yp)+f(p√xp−yp)=2f(x)g(y), | (Wr) |
f(p√xp+yp)−f(p√xp−yp)=2g(x)f(y), | (Trgf) |
f(p√xp+yp)−f(p√xp−yp)=2g(x)g(y), | (Trgg) |
f(p√xp+yp)−f(p√xp−yp)=2g(x)h(y). | (Trgh) |
We can confirm that each equation has a solution as follows: (Cr):f(x)=cos(xp), (Wr):f(x)=sin(xp),g(x)=cos(xp),(Trgf):f(x)=sin(xp), g(x)=cos(xp),(Trgg):f(x)=cos(xp), g(x)=isin(xp),(Trgh):f(x)=cos(xp), g(x)=sin(xp),h(x)=−sin(xp) .
In addition, the solution of each equation can also be found in perspective of the hyperbolic function, exponential function and p-power function, simultaneously.
Letting p = 1 in the above paragraph, we know that each original equation ((C), (W), (Tgf), (Tgg), (Tgh)) has the corresponding solution of the same form, respectively. They also are represented by the hyperbolic function, exponential function and p-power function, simultaneously.
Now let's consider the functional equations generated by the product of the above equations, then we obtain the target equations: p-power-radical functional equation (Sr) and (Sr)-type's equations (Srfg), (Srgf), (Srgg) and (Srgh).
1) (Sr) has a solution as the p-power function f(x)=xp:
f(p√x+y2)2−f(p√x−y2)2=(x+y2)2−(x−y2)2=xy=f(p√x)f(p√y). |
2) When (Sr) has a solution as the sine function, it also has simultaneously as an exponential solution as follows:
(12iei(p√xp+yp2)p−12ie−i(p√xp+yp2)p)2−(12iei(p√xp−yp2)p−12ie−i(p√xp−yp2)p)2=f(p√xp+yp2)2−f(p√xp−yp2)2=sin(xp+yp2)2−sin(xp−yp2)2=sin(xp)sin(yp)=f(x)f(y)=(eixp−e−ixp2i)(eiyp−e−iyp2i). |
3) When (Sr) has a solution as the hyperbolic sine function, it also has simultaneously as an exponential solution as follows:
(12e(p√xp+yp2)p−12e−(p√xp+yp2)p)2−(12e(p√xp−yp2)p−12e−(p√xp−yp2)p)2=f(p√xp+yp2)2−f(p√xp−yp2)2=sinh(xp+yp2)2−sinh(xp−yp2)2=sinh(xp)sinh(yp)=f(x)f(y)=(exp−e−xp2)(eyp−e−yp2). |
Although the mentioned all functional equations may have arisen from sine or cosine, as shown in the previous, they have solutions as the p-power, the hyperbolic and the exponential function, simultaneously. Hence, they can be considered as the p-power-radical, the p-power-radical exponential and the p-power-radical hyperbolic functional equation, simultaneously.
Letting p = 1 in the above items 1), 2) and 3), then (Sr) arrives (S). Hence, based on the above reasons, the Eq (S) well-known as the sine function equation can also be called as the p-power, the exponential and the hyperbolic functional equation, simultaneously.
In the following lemma, we find the forms of solutions of the p-power-radical functional equations (Srgh), (Srgg), (Srfg).
Lemma 1. If f,g,h:R→C satisfy (Srgh), then, as one of the solutions of (Srgh), f,g,h have the forms f(x)=cos(xp), g(x)=sin(xp) and h(x)=−sin(xp) for all x∈R.
Proof. For all x,y∈R,
(cosxp+yp2)2−(cosxp−yp2)2=f(p√xp+yp2)2−f(p√xp−yp2)2=g(x)h(y)=−sin(xp)sin(yp)=sin(xp)(−sin(yp)). |
In the next lemma, let's find an exponential solution for (Srfg).
Lemma 2. If f,g:R→C satisfy (Srfg), then, as the solutions of (Srfg), f,g have the following two forms
(i) f(x)=exp, g(x)=exp−e−xp for all x∈R,
(ii) f(x)=exp, g(x)=2sinh(xp) for all x∈R.
Proof. For all x,y∈R,
(exp+yp2)2−(exp−yp2)2=f(p√xp+yp2)2−f(p√xp−yp2)2=f(x)g(y)={(i)exp(eyp−e−yp)(ii)exp2sinh(yp). |
In the next lemma, let's find a hyperbolic and trigonometric solution for (Srgg).
Lemma 3. If f,g:R→C satisfy (Srgg), then, as the solutions of (Srgg), f,g have the following two forms
(i) f(x)=cosh(xp), g(x)=sinh(xp),
(ii) f(x)=cos(xp), g(x)=isin(xp).
Proof. For all x,y∈R,
(i)(coshxp+yp2)2−(coshxp−yp2)2=f(p√xp+yp2)2−f(p√xp−yp2)2=g(x)g(y)=sinhxpsinhyp. |
(ii)(cosxp+yp2)2−(cosxp−yp2)2=f(p√xp+yp2)2−f(p√xp−yp2)2=g(x)g(y)=isin(xp)isin(yp). |
In Section 3, we study the superstability of the p-power-radical sine functional equation (Sr) from an approximation of the p-power-radical functional equation (Srgh) related to (S).
Theorem 1. Assume that f,g,h:R⟶C satisfy the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)h(y)|≤φ(y) | (3.1) |
for all x,y∈R, where p is a positive odd integer.
Then, either g is bounded or h satisfies (Sr). Moreover, if g satisfies (Cr), then h and g satisfy p-power-radical equation (Trgf): = h(p√xp+yp)−h(p√xp−yp)=2g(x)h(y).
Proof. By putting x=p√2x and y=p√2y in (3.1), it is written equivalently as
|f(p√xp+yp)2−f(p√xp−yp)2−g(p√2x)h(p√2y)|≤φ(p√2y),∀x,y∈R. | (3.2) |
Assume that g is unbounded. Then, we can choose a sequence {xn} in R such that
0≠|g(p√2xn)|→∞,asn→∞. | (3.3) |
Taking x=xn in (3.2), we get
|f(p√xpn+yp)2−f(p√xpn−yp)2g(p√2xn)−h(p√2y)|≤φ(p√2y)|g(p√2xn)|, |
and by (3.3), we get
h(p√2y)=limn→∞f(p√xpn+yp)2−f(p√xpn−yp)2g(p√2xn). | (3.4) |
Replacing x by p√2xpn+xp and p√2xpn−xp in (3.1), we have
2φ(y)≥|g(p√2xpn+xp)h(y)−f(p√2xpn+xp+yp2)2+f(p√2xpn+xp−yp2)2|+|g(p√2xpn−xp)h(y)−f(p√2xpn−xp+yp2)2+f(p√2xpn−xp−yp2)2|≥|(g(p√2xpn+xp)+g(p√2xpn−xp))h(y)−(f(p√xpn+xp+yp2)2−f(p√xpn−xp+yp2)2)+(f(p√xpn+xp−yp2)2−f(p√xpn−xp−yp2)2)| | (3.5) |
for all x,y∈R and n∈N. Consequently,
2φ(y)|g(p√2xn)|≥|g(p√2xpn+xp)+g(p√2xpn−xp)g(p√2xn)h(y)−f(p√xpn+xp+yp2)2−f(p√xpn−xp+yp2)2g(p√2xn)+f(p√xpn+xp−yp2)2−f(p√xpn−xp−yp2)2g(p√2xn)| | (3.6) |
for all x,y∈R and n∈N. Taking n→∞ in (3.6) and using (3.3) and (3.4), we reach a conclusion that, for every x∈R, there exists the limit function
L1(x):=limn→∞g(p√2xpn+xp)+g(p√2xpn−xp)g(p√2xn), |
where L1:R→C satisfies the equation as even
h(p√xp+yp)−h(p√xp−yp)=L1(x)h(y),∀x,y∈R. | (3.7) |
From the definition of L1, we obtain the equality L1(0)=2, which, jointly with (3.7), indicates that h is odd. Keeping this in mind, through (3.7), we deduce the equality
h(p√xp+yp)2−h(p√xp−yp)2=[h(p√xp+yp)+h(p√xp−yp)]L1(x)h(y)=[h(p√2xp+yp)+h(p√2xp−yp)]h(y)=[h(p√yp+2xp)−h(p√yp−2xp)]h(y)=L1(y)h(p√2x)h(y). | (3.8) |
The oddness of h imposes it to vanish at 0. Putting x=y in (3.7), we conclude with the previous result that
h(p√2y)=h(y)L1(y). | (3.9) |
The (3.8) by (3.9) arrives to the equation
h(p√xp+yp)2−h(p√xp−yp)2=h(p√2x)h(p√2y), |
for all x,y∈R, which, with p√2-divisibility of R, states conclusively (Sr).
In addition, if g satisfies (Cr) and L1 forces 2g, then (3.7) forces that h and g satisfy (Trgf).
Theorem 2. Suppose that f,g,h:R⟶C satisfy
|f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)h(y)|≤φ(x),∀x,y∈R, | (3.10) |
which satisfies one of two cases g(0)=0, f(x)2=f(−x)2, where p is a positive odd integer.
Then, either h is bounded or g satisfies (Sr). In addition, if h satisfies (Cr), then g and h satisfy the p-power-radical Wilson type equation (Wr): = g(p√xp+yp)+g(p√xp−yp)=2g(x)h(y).
Proof. Let h be unbounded, then we can select a sequence {yn} in R such that h(p√2yn)|→∞ as n→∞. With a minor change of the steps shown in the start part of the proof in Theorem 1, we can get
g(p√2x)=limn→∞f(p√xp+ypn)2−f(p√xp−ypn)2h(p√2yn). | (3.11) |
Replacing y by p√yp+2ypn and p√−yp+2ypn in (3.10), the same procedure of (3.5) and (3.6) allows, with (3.11), use to argue the existence of a limit function
L2(y):=limn→∞h(p√yp+2ypn)+h(p√−yp+2ypn)h(p√2yn), |
where L2:R→C satisfies the equation
g(p√xp+yp)+g(p√xp−yp)=g(x)L2(y),∀x,y∈R. | (3.12) |
Hence, from the definition of L2, L2 is even and L2(0)=2.
Let's start with the case g(0)=0. Then it leads to the conclusion, by (3.12), that g is odd.
Putting y=x in (3.12), we obtain
g(p√2x)=g(x)L2(x),∀x,∈R. | (3.13) |
From (3.12), the oddness of g and (3.13), we obtain the equation
g(p√xp+yp)2−g(p√xp−yp)2=g(x)L2(y)[g(p√xp+yp)−g(p√xp−yp)]=g(x)[g(p√xp+2yp)−g(p√xp−2yp)]=g(x)[g(p√2yp+xp)+g(p√2yp−xp)]=g(x)g(p√2y)L2(x)=g(p√2x)g(p√2y) |
for all x,y∈R, which, with p√2-divisibility of R, states conclusively (Sr).
Second, let's consider the case f(x)2=f(−x)2. in this case, it is sufficient to show that g(0)=0.
Suppose that it is not the case. Then, without loss of generality, we may assume in following that g(0)=c (constant).
Taking x=0 in (3.10), from the above assumption, we get the inequality
|h(y)|≤φ(0)c,∀y∈R. |
The above inequality indicates that h is globally bounded, which is a contradiction due to the assumption of unboundedness. Therefore the claimed g(0)=0 holds, and the proof of the theorem is completed.
In addition, if h satisfies (Cr) and L2 forces 2h, then (3.12) forces that g and h satisfy (Wr): = g(p√xp+yp)+g(p√xp−yp)=2g(x)h(y).
The following corollary follows from Theorems 1 and 2, immediately.
Corollary 1. Assume that f,g,h:R→C satisfy the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)h(y)|≤min{φ(x),φ(y)},∀x,y∈R, | (3.14) |
where p is a positive odd integer.
Then
(i) either g is bounded or h satisfies (Sr). Moreover, if g satisfies (Cr), then h and g satisfy (Trgf): = h(p√xp+yp)−h(p√xp−yp)=g(x)h(y).
(ii) either h is bounded, or g satisfies (Sr) under g(0)=0 or f(x)2=f(−x)2. Moreover, if h satisfies (Cr), then g and h satisfy the Wilson equation (Wr): = g(p√xp+yp)+g(p√xp−yp)=2g(x)h(y).
In this section, as corollaries, we obtain the superstability of the p-power-radical sine functional equation (Sr) from an approximation of (Sr), and (Srgf), (Srfg) and (Srgg). Their proofs follow from Theorems 1 and 2, and Corollary 1.
Corollary 2. Assume that f,g:R⟶C satisfy the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)g(y)|≤{(i)φ(y)(ii)φ(x)(iii)min{φ(x),φ(y)}∀x,y∈R, |
where p is a positive odd integer.
Then, either g is bounded or g satisfies (Sr), respectively. In particular, the case (ii) holds under the condition g(0)=0 or f(x)2=f(−x)2.
Corollary 3. Assume that f,g:R→C satisfy the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)f(y)|≤φ(y) | (4.1) |
for all x,y∈R, where p is a positive odd integer.
Then, either g is bounded or f satisfy (Sr). Moreover, if g satisfies (Cr), f and g satisfy (Trgf): = f(p√xp+yp)−f(p√xp−yp)=2g(x)f(y).
Corollary 4. Assume that f,g:R→C satisfy the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)f(y)|≤φ(x),∀x,y∈R | (4.2) |
which satisfies one of the cases g(0)=0, f(x)2=f(−x)2, where p is a positive odd integer.
Then, either f is bounded or g satisfies (Sr). Additionally, if f satisfies (Cr), g and f satisfy (Wr): = g(p√xp+yp)+g(p√xp−yp)=2g(x)f(y).
Corollary 5. Assume that f,g:R→C satisfy the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)f(y)|≤min{φ(x),φ(y)} | (4.3) |
for all x,y∈R, where p is a positive odd integer.
Then
(i) either g is bounded or f and g satisfy (Sr), respectively. Additionally, if g satisfies (Cr), then f and g satisfy (Trgf): = f(p√xp+yp)−f(p√xp−yp)=2g(x)f(y);
(ii) either f is bounded or g satisfies (Sr). Additionally, if f satisfies (Cr), then g and f satisfy the Wilson equation (Wr): = g(p√xp+yp)+g(p√xp−yp)=2g(x)f(y).
Proof. It is sufficient to present that either g is bounded or g satisfies (S). The other cases follow from Corollaries 3 and 4, immediately.
The inequality (4.3) can also be presented equivalently as
|f(p√xp+yp)2−f(p√xp−yp)2−g(p√2x)f(p√2y)|≤min{φ(p√2x),φ(p√2y)},∀x,y∈R. | (4.4) |
First, if f is bounded, then y0∈R can be chosen such that f(p√2y0)≠0. From this y0 and (4.4), we get
|g(p√2x)|−|f(p√xp+yp0)2−f(p√xp−yp0)2f(p√2y0)|≤|f(p√xp+yp0)2−f(p√xp−yp0)2f(p√2y0)−g(p√2x)|≤min{φ(p√2x),φ(p√2y0)}f(p√2y0)≤φ(p√2y0)f(p√2y0). |
Thus, it implies that g is also bounded on R. Namely, since an unboundedness of g exacts it of f, let run along the step of Theorem 2.
The process of Theorem 2 gives us the limit (3.11), which, since f satisfies (Sr) by Theorem 1, validates
g(p√2x)=f(p√2x),∀x∈R. |
By the p√2-divisibility of R, we obtain g=f. Thus, it is true that g also satisfies (Sr).
Corollary 6. Assume that f,g:R⟶C satisfy the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−f(p√x)g(p√y)|≤φ(y), | (4.5) |
where p is a positive odd integer.
Then, either f is bounded or g satisfies (Sr). Additionally, if f satisfies (Cr), then g and f satisfy (Trgf): = g(p√xp+yp)−g(p√xp−yp)=2f(x)g(y).
Corollary 7. Assume that f,g:R⟶C satisfy the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−f(p√x)g(p√y)|≤φ(x), |
where p is a positive odd integer.
Then, either g is bounded or f satisfies (Sr) under one condition of the cases f(0)=0, f(x)2=f(−x)2. In addition, if g satisfies (Cr), then f and g satisfy the Wilson equation (Wr): = f(p√xp+yp)+f(p√xp−yp)=f(x)g(y).
Corollary 8. Assume that f,g:R⟶C satisfy the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−f(x)g(y)|≤min{φ(x),φ(y)} |
for all x,y∈R, where p is a positive odd integer.
Then
(i) either f is bounded or g satisfies (Sr). In addition, if f satisfies (Cr), then g and f satisfy (Trgf): = g(p√xp+yp)−g(p√xp−yp)=2f(x)g(y);
(ii) either g is bounded or f and g satisfy (Sr), respectively, under one condition of the cases f(0)=0, f(x)2=f(−x)2. In addition, if g satisfies (Cr), then f and g satisfy the Wilson equation (Wr): = f(p√xp+yp)+f(p√xp−yp)=2f(x)g(y).
As a corollary for all the obtained results, we obtain the superstability of the p-power-radical sine functional equation (Sr).
Corollary 9. Assume that f:R→C satisfies the inequality
|f(p√xp+yp2)2−f(p√xp−yp2)2−f(x)f(y)|≤{(i)φ(y),(ii)φ(x),(iii)min{φ(x),φ(y)}, |
where p is a positive odd integer.
Then, either f is bounded or f satisfies (Sr)
Proof. Replacing the functions g and h in Theorems 1 and 2 by f, in the case (ii), the assumption f(0)=0 or f(x)2=f(−x)2 can be eliminated (see [9, Theorem 5]).
Remark 1. (i) Applying φ(x)=φ(y)=ε for all results in Sections 3 and 4, then they yield the superstability results bounded by constant (Hyers-sense).
(ii) Applying 'p=1' to all the p-power-radical functional equations (Sr), (Srgf), (Srfg), (Srgg), (Srgh) in Sections 3 and 4, then they yield the superstability results for all (S)-type functional equations: (S), (Sgf), (Sfg), (Sgg), (Sgh).
In addition, for all results of the (S)-types obtained above, applying again (i) φ(x)=φ(y)=ε, then they yield the additional results (Hyers-sense) for (S)-types.
(iii) Many results obtained for the (S)-types and p-power-radical (Sr)-types in (i) and (ii) are found in Cholewa [16], Badora [8], Badora and Ger [9], Kannappan and Kim [10], Kim and Dragomir [11], Kim [18,22] and in papers [8,9,13,14,15].
All results in Sections 3 and 4 can be expanded to the stability on Banach algebras. The following theorem is based on Theorems 1 and 2, and Corollary 1. The remainder results also are represented as similar type as Theorem 3, respectively, their proofs will skip for the sake of brevity.
Theorem 3. Let (E,‖⋅‖) be a semisimple commutative Banach algebra. Assume that f,g,h:R⟶E satisfy the inequality
‖f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)h(y)‖≤{(i)φ(y),(ii)φ(x),(iii)min{φ(x),φ(y)}, |
where p is a positive odd integer.
Then, for an arbitrary linear multiplicative functional x∗∈E∗,
(i) either the superposition x∗∘g is bounded or h satisfies (Sr), In addition, if g satisfies (Cr), then h and g satisfy (Trgf): = h(p√xp+yp)−h(p√xp−yp)=2g(x)h(y);
(ii) either the superposition x∗∘h under the cases g(0)=0 or f(x)2=f(−x)2 is bounded or g satisfies (Sr). In addition, if h satisfies (Cr), then g and h satisfy the Wilson equation (Wr): = g(p√xp+yp)+g(p√xp−yp)=2g(x)h(y);
(iii) the above (i) and (ii) hold. In addition, if the superposition x∗∘g is unbounded, then g satisfies (Sr)
Proof. (i) Assume that (i) holds and fix arbitrarily a linear multiplicative functional x∗∈E. As is well known, we have ‖x∗‖=1, whence, for every x,y∈R, we have
φ(y)≥‖g(x)h(y)−f(p√xp+yp2)2+f(p√xp−yp2)2‖=sup‖y∗‖=1|y∗(g(x)h(y)−f(p√xp+yp2)2+f(p√xp−yp2)2)|≥|x∗(g(x))⋅x∗(h(y))−x∗(f(p√xp+yp2))+x∗(f(p√xp−yp2))|, |
which states that the superpositions x∗∘g and x∗∘h produce a solution of stability inequality (3.1) of Theorem 1. Since, by assumption, the superposition x∗∘g is unbounded, an appeal to Theorem 1 forces that the function x∗∘h is a solution of (Sr), that is,
(x∗∘h)(p√xp+yp2)2−(x∗∘h)(p√xp−yp2)2=(x∗∘h)(x)(x∗∘h)(y). | (5.1) |
In other presents, by the linear multiplicativity of x∗, for all x,y∈R, the difference DSr:R×R→E defined by
DSr(x,y):=h(p√xp+yp2)2−h(p√xp−yp2)2−h(x)h(y) |
falls into the kernel of x∗. Thus, in view of the unrestricted choice of x∗, we infer that
DSr(x,y)∈⋂{kerx∗:x∗ is a multiplicative member of E∗} |
for all x,y∈R. Since the space E is a semisimple, ⋂{kerx∗:x∗∈E∗}=0, which means that h satisfies the claimed Eq (Sr).
In addition, if g satisfies (Cr), then it is trivial that h and g satisfy h(p√xp+yp)−h(p√xp−yp)=2g(x)h(y).
(ii) By assumption, the superposition z∗∘h with g(0)=0 or f(x)2=f(−x)2 is unbounded, an appeal to Theorem 2 shows that the results hold.
The superposition x∗∘g satisfies (5.1), that is a solution of the Eq (Sr).
As in (i), a linear multiplicativity of x∗ and semisimplicity imply
g(p√xp+yp2)2−g(p√xp−yp2)2−g(x)g(y)∈⋂{kerx∗:x∗∈E∗}=0, |
which means that g satisfies (Sr). In addition, if h satisfies (Cr), then it is trivial that g and h satisfy g(p√xp+yp)+g(p√xp−yp)=2g(x)h(y).
(iii) It follows from the above (i) and (ii), and the additional case of (iii) holds by Corollary 1.
Corollary 10. Let (E,‖⋅‖) be a semisimple commutative Banach algebra. Assume that f,g:R⟶E satisfy the inequality
‖f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)g(y)‖≤{(i)φ(x),(ii)φ(y),(iii)min{φ(x),φ(y)}, |
where p is a positive odd integer.
For an arbitrary linear multiplicative functional x∗∈E∗, either the superposition x∗∘g is bounded or g satisfies (Sr). In particular, the case (ii) holds under the condition g(0)=0 or f(x)2=f(−x)2.
Corollary 11. Let (E,‖⋅‖) be a semisimple commutative Banach algebra. Assume that f,g:R⟶E satisfy the inequality
‖f(p√xp+yp2)2−f(p√xp−yp2)2−g(x)f(y)‖≤{(i)φ(x),(ii)φ(y),(iii)min{φ(x),φ(y)}, |
where p is a positive odd integer.
Then, for an arbitrary linear multiplicative functional x∗∈E∗,
(i) either the superposition x∗∘g is bounded or f satisfies (Sr), In addition, if g satisfies (Cr), then f and g satisfy (Trgf);
(ii) either the superposition x∗∘f under the cases g(0)=0 or f(x)2=f(−x)2 is bounded or g satisfies (Sr). In addition, if f satisfies (Cr), then g and f satisfy the Wilson equation (Wr);
(iii) the above (i) and (ii) hold. Also, additionally, if the superposition x∗∘g is unbounded, then g satisfies (Sr)
Corollary 12. Let (E,‖⋅‖) be a semisimple commutative Banach algebra. Assume that f,g:R⟶E satisfy the inequality
‖f(p√xp+yp2)2−f(p√xp−yp2)2−f(x)g(y)‖≤{(i)φ(x),(ii)φ(y),(iii)min{φ(x),φ(y)}, |
where p is a positive odd integer.
Then, for an arbitrary linear multiplicative functional x∗∈E∗,
(i) either the superposition x∗∘f is bounded or g satisfy (Sr). In addition, if f satisfies (Cr), then g and f satisfy (Trgf);
(ii) either the superposition x∗∘g under the cases g(0)=0 or f(x)2=f(−x)2 is bounded or f satisfies (Sr). In addition, if g satisfies (Cr), then f and g satisfy (Wr);
(iii) the above (i) and (ii) hold. Also, additionally, if the superposition x∗∘g is unbounded, then g satisfies (Sr),
Corollary 13. Let (E,‖⋅‖) be a semisimple commutative Banach algebra. Assume that f:R⟶E satisfies the inequality
‖f(p√xp+yp2)2−f(p√xp−yp2)2−f(x)f(y)‖≤{(i)φ(x),(ii)φ(y),(iii)min{φ(x),φ(y)}, |
where p is a positive odd integer.
For an arbitrary linear multiplicative functional x∗∈E∗, either the superposition x∗∘f is bounded or f satisfies (Sr).
Remark 2. Follow (i) and (ii) of Remark 1 for all results in Section 5, namely,
(i) Apply φ(x)=φ(y)=ε in all results.
(ii) Apply 'p=1' in all results. Next, apply (i) again in the results.
Then, a number of the results are found in the same papers in (iii) of Remark 1.
In this paper, we studied solutions and creating of the p-power-radical functional equations arisen simultaneously from the trigonometric functions, hyperbolic function, exponential function and p-radical function.
We investigated the superstability bounded by a function (Gǎvruta sense) for the p-power-radical sine functional equation (Sr) from an approximation of the p-power-radical functional equations (Srgh), and (Sr), (Srgf), (Srfg), (Srgg) with p is a positive odd integer. Furthermore, the obtained results extended to Banach algebras. As a result, we have improved the previous stability results for (S)-type functional equations: (S), (Sgf), (Sfg), (Sgg), (Sgh) to that of the p-power-radical equations: (Sr), (Srgf), (Srfg), (Srgg), (Srgh).
The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.
This study was supported by research fund from Chosun University, 2023.
The authors declare there are no conflicts of interest.
[1] |
L. I. Rudin, S. Osher, E. Fatemi, Nonlinear total variation based noise removal algorithms, Phys. D, 60 (1992), 259–268. doi: 10.1016/0167-2789(92)90242-F
![]() |
[2] | K. Bredies, H. P. Sun, Preconditioned Douglas-Rachford algorithms for TV and TGV-regularized variational imaging problems, J. Math. Imaging Vis., 53 (2015), 317–344. |
[3] |
S. Wang, T. Z. Huang, J. Liu, X. G. Lv, An alternating iterative algorithm for image deblurring and denoising problems, Commun. Nonlinear Sci. Numer. Simul., 19 (2014), 617–626. doi: 10.1016/j.cnsns.2013.07.004
![]() |
[4] |
Y. L. Wang, J. F. Yang, W. Yin, Y. Zhang, A new alternating minimization algorithm for total variation image reconstruction, SIAM J. Imaging Sci., 1 (2008), 248–272. doi: 10.1137/080724265
![]() |
[5] |
A. Chambolle, An algorithm for total variation minimization and applications, J. Math. Imaging Vis., 20 (2004), 89–97. doi: 10.1023/B:JMIV.0000011321.19549.88
![]() |
[6] |
A. Beck, M. Teboulle, Fast gradient-based algorithm for constrained total variation denoising and deblurring problems, IEEE Trans. Image Process., 18 (2009), 2419–2434. doi: 10.1109/TIP.2009.2028250
![]() |
[7] | G. A. Tsihrintzis, Statistical modeling and receiver design for multi-user communication networks, In: R. J. Adler, R. E. Feldman, M. S. Taqqu, A practical guide to heavy tails: Statistical techniques and applications, Birkhäuser, 1998. |
[8] | B. Kosko, Noise, New York: Viking Press, 2006. |
[9] |
T. Pander, New polynomial approach to myriad filter computation, Signal Process., 90 (2010), 1991–2001. doi: 10.1016/j.sigpro.2010.01.001
![]() |
[10] |
Y. C. Chang, S. R. Kadaba, P. C. Doerschuk, S. B. Gelfand, Image restoration using recursive Markov random field models driven by Cauchy distributed noise, IEEE Signal Process. Lett., 8 (2001), 65–66. doi: 10.1109/97.905941
![]() |
[11] |
A. Loza, D. Bull, N. Canagarajah, A. Achim, Non-Gaussian model-based fusion of noisy images in the wavelet domain, Comput. Vis. Image Und., 114 (2010), 54–65. doi: 10.1016/j.cviu.2009.09.002
![]() |
[12] |
T. Wan, N. Canagarajah, A. Achim, Segmentation of noisy colour images using Cauchy distribution in the complex wavelet domain, IET Image Process., 5 (2011), 159–170. doi: 10.1049/iet-ipr.2009.0300
![]() |
[13] |
F. Sciacchitano, Y. Q. Dong, T. Y. Zeng, Variational approach for restoring blurred images with Cauchy noise, SIAM J. Imaging Sci., 8 (2015), 1894–1922. doi: 10.1137/140997816
![]() |
[14] | A. C. Bovik, Handbook of image and video processing, New York: Academic Press, 2000. |
[15] |
J. J. Mei, Y. Dong, T. Z. Huang, W. Yin, Cauchy noise removal by non-convex ADMM with convergence guarantees, J. Sci. Comput., 74 (2018), 743–766. doi: 10.1007/s10915-017-0460-5
![]() |
[16] | K. Bredies, K. Kunisch, T. Pock, Total generalized variation, SIAM J. Imaging Sci., 3 (2010), 492–526. |
[17] |
F. Knoll, K. Bredies, T. Pock, R. Stollberger, Second order total generalized variation (TGV) for MRI, Magn. Reson. Med., 65 (2011), 480–491. doi: 10.1002/mrm.22595
![]() |
[18] |
K. Bredies, K. Kunisch, T. Valkonen, Properties of L1-TGV2: The one-dimensional case, J. Math. Anal. Appl., 398 (2013), 438–454. doi: 10.1016/j.jmaa.2012.08.053
![]() |
[19] | Q. X. Zhong, C. S. Wu, Q. L. Shu, R. W. Liu, Spatially adaptive total generalized variation-regularized image deblurring with impulse noise, J. Electron. Imaging, 27 (2018), 053006. |
[20] |
W. Q. Lu, J. M. Duan, Z. W. Qiu, Z. K. Pan, R. W. Liu, L. Bai, Implementation of high-order variational models made easy for image processing, Math. Methods Appl. Sci., 39 (2016), 4208–4233. doi: 10.1002/mma.3858
![]() |
[21] |
R. W. Liu, L. Shi, W. H. Huang, J. Xu, S. C. H. Yu, D. F. Wang, Generalized total variation-based MRI Rician denoising model with spatially adaptive regularization parameters, Mag. Reson. Imaging, 32 (2014), 702–720. doi: 10.1016/j.mri.2014.03.004
![]() |
[22] |
R. W. Liu, L. Shi, S. C. H. Yu, D. F. Wang, Box-constrained second-order total generalized variation minimization with a combined L1, 2 data-fidelity term for image reconstruction, J. Electron. Imaging, 24 (2015), 033026. doi: 10.1117/1.JEI.24.3.033026
![]() |
[23] |
Y. Dong, T. Zeng, A convex variational model for restoring blurred images with multiplicative noise, SIAM J. Imaging Sci., 6 (2013), 1598–1625. doi: 10.1137/120870621
![]() |
[24] |
G. Aubert, J. F. Aujol, A variational approach to removing multiplicative noise, SIAM J. Appl. Math., 68 (2008), 925–946. doi: 10.1137/060671814
![]() |
[25] | K. Bredies, Recovering piecewise smooth multichannel images by minimization of convex functionals with total generalized variation penalty, In: A. Bruhn, T. Pock, X. C. Tai, Efficient algorithms for global optimization methods in computer vision, Lecture Notes in Computer Science, Berlin: Springer, 8293 (2014), 44–77. |
[26] |
L. F. Bai, A new non-convex approach for image restoration with Gamma noise, Comput. Math. Appl., 77 (2019), 2627–2639. doi: 10.1016/j.camwa.2018.12.045
![]() |
[27] | W. Feller, An introduction to probability theory and its applications, John Wiley & Sons, 2008. |
[28] | D. P. Bertsekas, A. Nedić, A. E. Ozdaglar, Convex analysis and optimization, Athena Scientific, 2003. |
[29] |
R. Bergmann, A. Weinmann, A second-order TV-type approach for inpainting and denoising higher dimensional combined cyclic and vector space data, J. Math. Imaging Vis., 55 (2016), 401–427. doi: 10.1007/s10851-015-0627-3
![]() |
[30] | N. Parikh, S. Boyd, Proximal algorithms, Found. Trends Optim., 1 (2014), 127–239. |
[31] |
A. Chambolle, T. Pock, A first-order primal-dual algorithm for convex problems with applications to imaging, J. Math. Imaging Vis., 40 (2011), 120–145. doi: 10.1007/s10851-010-0251-1
![]() |
[32] |
W. H. Guo, J. Qin, W. T. Yin, A new detail-preserving regularity scheme, SIAM J. Imaging Sci., 7 (2014), 1309–1334. doi: 10.1137/120904263
![]() |
[33] | N. Jacobson, Basic algebra II, San Francisco: Freeman Company, 1980. |
[34] |
C. Byrne, A unified treatment of some iterative algorithms in signal processing and image reconstruction, Inverse Prob., 20 (2004), 103–120. doi: 10.1088/0266-5611/20/1/006
![]() |
[35] |
P. L. Combettes, V. R. Wajs, Signal recovery by proximal forward-backward splitting, Multiscale Model. Simul., 4 (2005), 1168–1200. doi: 10.1137/050626090
![]() |
[36] | Y. W. Wen, M. K. Ng, W. K. Ching, Iterative algorithms based on decoupling of deblurring and denoising for image restoration, SIAM J. Sci. Comput., 30 (2007), 2655–2674. |
[37] |
Y. M. Huang, M. K. Ng, Y. W. Wen, A fast total variation minimization method for image restoration, Multiscale Model. Simul., 7 (2008), 774–795. doi: 10.1137/070703533
![]() |
[38] | C. L. Byrne, Applied iterative methods, New York: AK Peters/CRC Press, 2007. |
[39] | B. R. Frieden, A new restoring algorithm for the preferential enhancement of edge gradients, J. Opt. Soc. Am., 66 (1976), 116–123. |
1. | Sylvia Kirchengast, Beda Hartmann, Recent Lifestyle Parameters Are Associated with Increasing Caesarean Section Rates among Singleton Term Births in Austria, 2018, 16, 1660-4601, 14, 10.3390/ijerph16010014 | |
2. | Imnameren Longkumer, Naorem Kiranmala Devi, Benrithung Murry, Kallur Nava Saraswathy, Differential risk factors and morbidity/mortality pattern in type 2 diabetes: A study among two Mendelian populations with different ancestry (India), 2020, 14, 18714021, 1769, 10.1016/j.dsx.2020.09.006 |