These are chat archives for fiji/fiji

9th
Nov 2017
Varun Kapoor
@kapoorlab
Nov 09 2017 12:48
Hey Community people, I want to implement this algorithm in Fiji https://hal.archives-ouvertes.fr/inria-00184490/document, in case anyone has already worked on detecting circles in an image using Ransac, kindly let me know. Greetings from Paris!
Jean-Yves Tinevez
@tinevez
Nov 09 2017 13:09
Working with @gselzer and @ctrueden on Hough circle transform, but this is as far as it goes...
Varun Kapoor
@kapoorlab
Nov 09 2017 13:11
@tinevez Hough circle transform is too computationally demanding, I would start with Ransac implementation then by extending Saalfelds mpi-cbg package for function fits
Jean-Yves Tinevez
@tinevez
Nov 09 2017 13:12
oki. Will you do it in imglib2?
Varun Kapoor
@kapoorlab
Nov 09 2017 13:13
Yeah, me and Stephan extended that to include Ransac fits for polynomial of any degree and now I can do an extension of that to include circular/ elliptical curves.
Stephan Saalfeld
@axtimwalde
Nov 09 2017 16:30
@kapoorlab n-d polynomials of any degree?
Varun Kapoor
@kapoorlab
Nov 09 2017 18:04
n-2d polynomials of any degree, I do not think it is for dimension greater than 2 but any degree polynomial in two dimension has been implemented: https://github.com/PreibischLab/FunctionFit
(then we could directly use them in our alignment pipelines, and I would love to do this :))
(or wrap, not sure what the best and easiest mode is)
Stephan Saalfeld
@axtimwalde
Nov 09 2017 19:38
the only interesting part is probably how the coefficients are ordered, my order is this https://github.com/axtimwalde/mpicbg/blob/master/mpicbg/src/main/java/mpicbg/models/PolynomialTransform2D.java#L41-L60 following the LONI 'standard' ;)
Stephan Preibisch
@StephanPreibisch
Nov 09 2017 20:32
Hi @axtimwalde, not sure if you remember, we discussed this about two years ago after I implemented linear and quadratic fits extending mpi-cbg with that
Now @kapoorlab wrote a polynomial implementation of it
We can discuss maybe in Janelia how to merge our two extensions for function fits
should be easy
@kapoorlab: @tpietzsch and myself wrote a circle-RANSAC fit some time ago
Varun Kapoor
@kapoorlab
Nov 09 2017 20:36
@StephanPreibisch Ah I c, good you told me that I was planning to start writing it tomorrow, github link for circle-Ransac?
Stephan Preibisch
@StephanPreibisch
Nov 09 2017 20:36
I have to find it ... It also uses our extension of the mpicbg library for function fits
for 2d and 3d
I mean, it is rather simple
2d/3d ellipsoid would be harder and we didn't do it, maybe you can try that?
@axtimwalde I also implemented regularized functions, super cool :)
Varun Kapoor
@kapoorlab
Nov 09 2017 20:39
Sure, should be doable, I can integrate that into the function fits as well.
Stephan Saalfeld
@axtimwalde
Nov 09 2017 22:47
@StephanPreibisch you mean the interpolated functions?