Multi-Gauss Fit

[Burni]

Greetings,

I made a thread some days ago about finding peaks in a plot profile.
What I want to do now is use these peaks as an estimation for the parameters of a Gaussian Fit.
I will be working with a Profile like this:



With the help of the people in the old thread, I can now find the positions of the absolute peaks very reliably.
The problem I have now is the fact that ImageJ does not seem to support multi-fitting.
Even when I tried the fitting tool for 'User-defined functions', it only allows 6 parameters. But for a multi-gauss fit with 2 peaks, I would already need 7.
Does anybody perhaps know of an addon that supports more options for fitting? Specifically a gaussian multi-fit would help me greatly.

Burni

[Herbie-3]

Burni,

may I remind you that ImageJ is for image processing. Although it
supplies a number of routines that are suitable for 1D-signal
processing, there are better tools available for this kind of tasks.

Furthermore I don't think that at least some of the peaks should be
fitted by a Gaussian. Is there any a priori reason for your assumption?

Best

Herbie

:::::::::::::::::::::::::::::::
On 23.03.14 14:06, Burni wrote:

> Greetings,
>
> I made a thread some days ago about finding peaks in a plot profile.
> What I want to do now is use these peaks as an estimation for the parameters
> of a Gaussian Fit.
> I will be working with a Profile like this:
>
> <http://imagej.1557.x6.nabble.com/file/n5007047/PlotProfile.png>
>
> With the help of the people in the old thread, I can now find the positions
> of the absolute peaks very reliably.
> The problem I have now is the fact that ImageJ does not seem to support
> multi-fitting.
> Even when I tried the fitting tool for 'User-defined functions', it only
> allows 6 parameters. But for a multi-gauss fit, I would already need 7.
> Does anybody perhaps know of an addon that supports more options for
> fitting? Specifically a gaussian multi-fit would help me greatly.
>
> Burni
>
>
>
> --
> View this message in context: http://imagej.1557.x6.nabble.com/Multi-Gauss-Fit-tp5007047.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Jirka]

Hello,
I have in CMP-BIA tools Mixture Model fitting using kmeans or EM, but it
just api functionality not independent plugin.

Have nice day, JIrka

--
Best regards, Jiri Borovec
----------------------------------------------------------------------------------------------------
Jiří Borovec <[e-mail]>
PhD student at CMP CTU, ISC member
http://cmp.felk.cvut.cz/~borovji3


On 23 March 2014 14:06, Burni <[e-mail]> wrote:

> Greetings,
>
> I made a thread some days ago about finding peaks in a plot profile.
> What I want to do now is use these peaks as an estimation for the
> parameters
> of a Gaussian Fit.
> I will be working with a Profile like this:
>
> <http://imagej.1557.x6.nabble.com/file/n5007047/PlotProfile.png>
>
> With the help of the people in the old thread, I can now find the positions
> of the absolute peaks very reliably.
> The problem I have now is the fact that ImageJ does not seem to support
> multi-fitting.
> Even when I tried the fitting tool for 'User-defined functions', it only
> allows 6 parameters. But for a multi-gauss fit, I would already need 7.
> Does anybody perhaps know of an addon that supports more options for
> fitting? Specifically a gaussian multi-fit would help me greatly.
>
> Burni
>
>
>
> --
> View this message in context:
> http://imagej.1557.x6.nabble.com/Multi-Gauss-Fit-tp5007047.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Burni]

Hello Herbie,

you are probably right, ImageJ may not be optimal for this kind of task.
I experimented with a number of fits. Gaussian fits should work best in theory, but sometimes Lorentz fits work better. You could presumably fit some peaks even better using a third function.
From my experimentation however, by far most peaks are fitted well enough with either Gauss or Lorentz, so I want to limit myself to these two for now.
But in any case, I agree that I should use ImageJ mainly for the image processing. I'll be looking for a better alternative.

Thanks for your response,

Burni

[Burni]

Hello Jiri,

thanks for your response, I will look into that.

[Herbie-3]

Dear experts,

from time to time, and in a certain case only recently, a fully "ImageJ
Macro"-recordable "ImageJ"-PlugIn appears to be needed that performs the
one-dimensional Fourier-Transformation. Meanwhile I've coded such a
PlugIn (free but not open source). Bundled with an introductory ReadMe,
it can be downloaded from

<http://www.gluender.de/Miscellanea/MiscTexts/UtilitiesText.html>

The PlugIn contains a classic implementation of the FFT-algorithm and as
such it requires a straight line selection (in an image ) of a length
measuring a power of two. It can either output the Power Spectrum, log10
Power Spectrum, Magnitude- & Phase-Spectrum, or Real- &
Imaginary-Spectrum. The spectra for positive frequencies are either
saved as Tab-delimited Text-files, displayed as "ImageJ"-Plots, or they
replace the pixel-values of the line selection.

A wrapper "ImageJ"-Macro for the "1D FFT"-processing of images or stacks
is available on request.

Best regards

Herbie

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[bnorthan]

Hi Herbie

I am curious any reason your plugin is "free" but isn't open source??

Do you use GitHub or anything like that??

Brian




On Fri, Apr 4, 2014 at 10:54 AM, Herbie <[hidden email]> wrote:

> Dear experts,
>
> from time to time, and in a certain case only recently, a fully "ImageJ
> Macro"-recordable "ImageJ"-PlugIn appears to be needed that performs the
> one-dimensional Fourier-Transformation. Meanwhile I've coded such a PlugIn
> (free but not open source). Bundled with an introductory ReadMe, it can be
> downloaded from
>
> <http://www.gluender.de/Miscellanea/MiscTexts/UtilitiesText.html>
>
> The PlugIn contains a classic implementation of the FFT-algorithm and as
> such it requires a straight line selection (in an image ) of a length
> measuring a power of two. It can either output the Power Spectrum, log10
> Power Spectrum, Magnitude- & Phase-Spectrum, or Real- & Imaginary-Spectrum.
> The spectra for positive frequencies are either saved as Tab-delimited
> Text-files, displayed as "ImageJ"-Plots, or they replace the pixel-values
> of the line selection.
>
> A wrapper "ImageJ"-Macro for the "1D FFT"-processing of images or stacks
> is available on request.
>
> Best regards
>
> Herbie
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html