Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The functional $E$ is defined on some real valued functions $\phi$ defined on $\Omega\subseteq\mathbb R^2$ ($R^{2}\rightarrow R$)and $E(\phi)$ is a real number. $\phi$ is differentiable. My equation is as follows: $E\left(\phi\right)=\int_{\Omega}\left|\left(g\left(\vec{x}\right)-A\right)\right|^{2}\left(1-H(\phi\left(\vec{x}\right))\right)d\vec{x} +\mu\int_{\Omega}H(\phi\left(\vec{x})\right)d\vec{x} +\nu\int_{\Omega}\delta\left(\phi\left(\vec{x}\right)\right)\left|\nabla\phi\left(\vec{x}\right)\right|d\vec{x}$

where $g$ is a known function $R^{2}\rightarrow R$, and A is a know scalar, $H$ is a Heaviside function,i.e. $H\left(\phi\right)=\frac{1}{2}+\frac{1}{\pi}\arctan\left(\phi\right)$ (I choose this kind of approximation to smear out $H$) and $\delta$ is a Delta function(I also smear out $\delta=\frac{\partial{H}}{\partial{\phi}}$), I want to know whether $E$ is continuous. I specify the norm is $\|\|_2$. Since $E(\phi)\in R$, $\|E(\phi)-E(\phi_{0})\|_{2}=|E(\phi)-E(\phi_{0})|$ And I am not sure whether my derivation is right.

Is this right???

For any $\varepsilon>0$, given:

$\underset{x\epsilon\Omega}{max}|\phi\left(x\right)-\phi_{0}\left(x\right)|<\frac{\varepsilon}{\int|\frac{\left(\mu-N\left(x\right)\right)}{\pi}|\, dx}$


  1. $E\left(\phi\right)-E\left(\phi_{0}\right)=\int_{\Omega}\left|\left(g\left(\vec{x}\right)-A\right)\right|^{2}\left(H(\phi_{0}\left(\vec{x}\right))-H(\phi\left(\vec{x}\right))\right)d\vec{x} +\mu\int_{\Omega}\left(H(\phi\left(\vec{x})\right)-H(\phi_{0}\left(\vec{x}\right))\right)d\vec{x} +\nu\int_{\Omega}\left(\delta\left(\phi\left(\vec{x}\right)\right)\left|\nabla\phi\left(\vec{x}\right)\right|-\delta\left(\phi_{0}\left(\vec{x}\right)\right)\left|\nabla\phi_{0}\left(\vec{x}\right)\right|\right)d\vec{x}$

  2. $E\left(\phi\right)-E\left(\phi_{0}\right)=\int_{\Omega}\left(\left|\left(g\left(\vec{x}\right)-A\right)\right|^{2}-\mu\right)\left(H(\phi_{0}\left(\vec{x}\right))-H(\phi\left(\vec{x}\right))\right)d\vec{x} +\nu\int_{\Omega}\left(\delta\left(\phi\left(\vec{x}\right)\right)\left|\nabla\phi\left(\vec{x}\right)\right|-\delta\left(\phi_{0}\left(\vec{x}\right)\right)\left|\nabla\phi_{0}\left(\vec{x}\right)\right|\right)d\vec{x}$

  3. $E\left(\phi\right)-E\left(\phi_{0}\right)<\int_{\Omega}\left(\left|\left(g\left(\vec{x}\right)-A\right)\right|^{2}-\mu\right)max\left(H(\phi_{0})-H(\phi)\right)d\vec{x} +\nu\int_{\Omega}\left(\delta\left(\phi\left(\vec{x}\right)\right)\left|\nabla\phi\left(\vec{x}\right)\right|-\delta\left(\phi_{0}\left(\vec{x}\right)\right)\left|\nabla\phi_{0}\left(\vec{x}\right)\right|\right)d\vec{x}$

And let $H$ is signed distance function. So$\left|\nabla\phi_{0}\left(\vec{x}\right)\right|\approx1$ for all $\vec{x}\in\Omega$. Then equation 3 becomes:

Four. $E\left(\phi\right)-E\left(\phi_{0}\right)<max\left(H(\phi_{0})-H(\phi)\right)\int_{\Omega}\left(\left|\left(g\left(\vec{x}\right)-A\right)\right|^{2}-\mu\right)d\vec{x} +\nu\int_{\Omega}\left(\delta\left(\phi\left(\vec{x}\right)\right)-\delta\left(\phi_{0}\left(\vec{x}\right)\right)\right)d\vec{x}$

$$max(\delta(\phi)-\delta(\phi_{0})) = (\delta(\phi(\vec{x^{*}}))-\delta(\phi_{0}(\vec{x^{*}})))$$ $$\vec{x^{*}}\in{\left\{x|max(\delta(\phi(\vec{x}))-\delta(\phi_{0}\vec{x}))for\,\, all\, x\in\Omega\right\}}$$

Five. $E\left(\phi\right)-E\left(\phi_{0}\right)<max\left(H(\phi_{0})-H(\phi)\right)\int_{\Omega}\left(\left|\left(g\left(\vec{x}\right)-A\right)\right|^{2}-\mu\right)d\vec{x} +\nu max(\delta(\phi)-\delta(\phi_{0}))\int_{\Omega}d\vec{x}$

Make $max(\delta(\phi)-\delta(\phi_{0}))=\frac{\varepsilon}{2\nu\int_{\Omega}d\vec{x}}$ and\,\, $max(H(\phi)-H(\phi_{0}))=\frac{\varepsilon}{2\int_{\Omega}\left(\left|\left(g\left(\vec{x}\right)-A\right)\right|^{2}-\mu\right)d\vec{x}}$

$E\left(\phi\right)-E\left(\phi_{0}\right)<\varepsilon$ So $E$ is continuous. I am not sure whether I can derive like this since I am confused little about $\phi$. $\phi$ is a function $R^2\rightarrow R$ and so $H$ is $R\rightarrow R$.

I really appreciate if you could tell my whether this is right or how can I to deriive it.

share|cite|improve this question
First: continuity depends on the notion of convergence for the functions $\phi$ (more precisely, on the topology on the space of functions $\phi$). Which one do you use? Second: What is $H$ precisely? Is is a function of the real line? Is it the discontinuous Heaviside or a differentiable approximation? Third: Is this related to "active contours without edges"? – Dirk Feb 10 '12 at 8:11
Thank you, Dirk. H is defined as above question. This function is a energy function which I want to minimize. Yes, this is related to "active contours without edges". – user18481 Feb 10 '12 at 9:19

Since the question leaves much to be explained, let us try to do some reconstruction (note that this mainly expands on the first part of @Dirk's comment). The functional $E$ is defined on some real valued functions $\phi$ defined on $\Omega\subseteq\mathbb R^2$ and $E(\phi)$ is a real number. For $E(\phi)$ to exist, it is necessary that $\phi$ is measurable, and even differentiable since the last integral involves the gradient of $\phi$.

To summarize, one considers a functional $E:\mathfrak F\to\mathbb R$, $\phi\mapsto E(\phi)$, defined on some unspecified subspace $\mathfrak F$ of $C^1(\Omega,,\mathbb R)$. The continuity of $E$ at $\phi$ in $\mathfrak F$ would be the statement that $E(\psi)$ is close to $E(\phi)$ for every $\psi$ in $\mathfrak F$ close to $\phi$. Closeness in such settings is often (but not always) quantified by a norm, say $\|\ \|_{\mathfrak F}$, and the continuity of $E$ at $\phi$ in $\mathfrak F$ would then read as follows: $$ \forall\varepsilon\gt0,\quad\exists\alpha\gt0,\quad\forall\psi\in\mathfrak F,\quad \|\phi-\psi\|_{\mathfrak F}\leqslant\alpha\implies|E(\psi)-E(\phi)|\leqslant\varepsilon. $$ Hence, what is also missing at this stage is a definition of $\|\ \|_{\mathfrak F}$. One can guess that $\phi$ or integrals of $\phi$ could be involved, but also $\nabla\phi$ or some of its integrals. In any case, to specify the space $\mathfrak F$ and the norm $\|\ \|_{\mathfrak F}$ is mandatory for the question to make sense.

share|cite|improve this answer
Thank you so much for your answer. The assumption is totally right. I will reedit my question based on your answer after I fix the problem of my exploreer. – user18481 Feb 14 '12 at 2:43
You seem to have made up your mind about the norm and chosen $\|\ \|_\mathfrak F=\|\ \|_2$ (the only available definition of a norm anywhere in your post, so far). This choice is hopeless, as explained in my post, since one needs a control on the gradient as well. – Did Mar 4 '12 at 9:53
Correction: in the new newest version, $\|\ \|_\mathfrak F$ becomes $\|\ \|_\infty$ on $\Omega$. Same remark as before. – Did Mar 4 '12 at 11:33
thank you again for your comment. No, actually, I use absolute value as the norm. I want to know whether $E$ is continuous since I want to calculate its minimal and I really did. I use iteration to solve the mininal. I mean I calculate the minianal iteratively. I update $\phi$ each time to try to maintain a signed distance function. If I see this kind of update is tiny, can I know whether $E$ is continuous proximately? – user18481 Mar 4 '12 at 11:36
Correction:I use iteration to solve the minimum, not mininal. Sorry for that. – user18481 Mar 4 '12 at 11:51

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.