How to test algorithms

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

How to test algorithms

Adrián Villalba
Dear colleagues,

How can i test different algorithms simoultaneously in one step?

For example, if i want to test which algorithm of threshold adjusting fits
better in my pic i have to do that manually one by one. Is it possible to
test all at the same time?

Thanks

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: How to test algorithms

ctrueden
Hi Adrián,

> if i want to test which algorithm of threshold adjusting fits better
> in my pic i have to do that manually one by one. Is it possible to
> test all at the same time?

For thresholding specifically, the Fiji distribution of ImageJ includes the
Image > Adjust > Auto Threshold command, which lets you "Try All",
producing a mosaic.

For image filtering algorithms in general, one option is to use ImageJ2's
notebook.mosaic method. Here is an example Groovy script which performs
N-dimensional Gaussian blur with various settings, and displays the results
side-by-side in a 2x2 grid:

  // @ImageJ ij
  // @Dataset image
  // @OUTPUT Object mosaic

  // Smudge it up horizontally.
  double[] horizSigmas = [8, 0, 0]
  horizGauss = ij.op().filter().gauss(image, horizSigmas)

  // And now vertically.
  double[] vertSigmas = [0, 8, 0]
  vertGauss = ij.op().filter().gauss(image, vertSigmas)

  // We can also blur the channels.
  double[] channelSigmas = [0, 0, 1]
  channelGauss = ij.op().filter().gauss(image, channelSigmas)

  int[] grid = [2, 2]
  mosaic = ij.notebook().mosaic(grid, image, horizGauss, vertGauss,
channelGauss)

And here is sample output from a downsampled Fluorescent Cells sample:

Of course, you can also just show the resulting images in separate windows.

Regards,
Curtis

--
Curtis Rueden
LOCI software architect - https://loci.wisc.edu/software
ImageJ2 lead, Fiji maintainer - https://imagej.net/User:Rueden
Did you know ImageJ has a forum? http://forum.imagej.net/


On Sun, Apr 9, 2017 at 3:48 PM, Adrián Villalba <[hidden email]> wrote:

> Dear colleagues,
>
> How can i test different algorithms simoultaneously in one step?
>
> For example, if i want to test which algorithm of threshold adjusting fits
> better in my pic i have to do that manually one by one. Is it possible to
> test all at the same time?
>
> Thanks
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

fluorescent-cells-mosaic.png (1M) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: How to test algorithms

gankaku
In reply to this post by Adrián Villalba
Dear Adrián,

if you mean compare with test and you use the Fiji distribution of ImageJ,
then you can have a look at the BioVoxxel Toolbox which offers the
"Threshold Check" comparing the automatic thresholds Curtis already
mentioned and it offers to do that in a visual/qualitative as well as a
semi-quantitative way. There is currently no implementation to test
user-chosen manual thresholds, but I can have a look in my old
repositories. I might still have something like this.
If you wnat to compare image filters you can equivalently have a look at
the "Filter Check" which gives you the chance to run multiple radii for one
specific filter making a comparison somewhat faster.

For more information, please have a look here:
http://imagej.net/BioVoxxel_Toolbox

Kind regards,
Jan

2017-04-09 22:48 GMT+02:00 Adrián Villalba <[hidden email]>:

> Dear colleagues,
>
> How can i test different algorithms simoultaneously in one step?
>
> For example, if i want to test which algorithm of threshold adjusting fits
> better in my pic i have to do that manually one by one. Is it possible to
> test all at the same time?
>
> Thanks
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>



--

CEO: Dr. Jan Brocher
phone:  +49 (0)621 62 979 313
mobile: +49 (0)176 705 746 81
e-mail: [hidden email]
info: [hidden email]
inquiries: [hidden email]
web: www.biovoxxel.de

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