These are chat archives for fiji/fiji

28th
Nov 2017
Varun Kapoor
@kapoorlab
Nov 28 2017 13:57
Hey so updates about the ellipse fitter: It can now detect multiple instances of ellipses in an image and I also implemented the point of intersection part of David Eberly`s follow up manuscript to enable detection of the point of intersections of the ellipses. Some examples: https://twitter.com/EntraCod/status/935507628456861696 and https://twitter.com/EntraCod/status/935204427123785729.
Curtis Rueden
@ctrueden
Nov 28 2017 15:11
@kapoorlab Cool! Could you please link to the code again?
Curtis Rueden
@ctrueden
Nov 28 2017 17:41
@kapoorlab Any chance those ellipsoid classes can be reconciled more with imglib2-roi's shape-roi branch? /cc @awalter17
Varun Kapoor
@kapoorlab
Nov 28 2017 18:34
@ctrueden I am actually not familiar with that repo, but what I want to do with my Ellipsoid and GeneralEllipsoid is to determine them in an image using the former and determine the points of intersection using the latter (this is how they appear on the Eberly`s text, apparently it is easier to get intersections using the quadratic form of ellipse than the normal form, https://www.geometrictools.com/Documentation/IntersectionOfEllipses.pdf ). Intersections have been worked out for 2D case only. So if I reconcile the ellipsoid (standard form of ellipse) and Generalellipsoid (Quadratic form of ellipse) will I be able to keep all the functionality I currently have? At the moment I can draw any intersecting ellipses, find them in an image without making spurious detections and get the point of intersections and draw tangents at that point. If we can preserve this functionality then sure reconciliation should be done after I have made myself familiar with the repo you mentioned. Thanks.
Current functionality in one image: https://twitter.com/EntraCod/status/935578345164730368
Curtis Rueden
@ctrueden
Nov 28 2017 18:43
@kapoorlab Thanks for the explanation. @awalter17 if you could please read over the above, and offer any comments, I'd appreciate it very much. There might be something here that could move upstream—or maybe everything is fine the way it is.
Alison Walter
@awalter17
Nov 28 2017 19:08
@kapoorlab I need to run to class soon, but I think you should be able to reconcile your ellipsoid classes with mine in imglib2-roi on the shape-rois branch. I’d have to look into your work further to say definitively, but from what I saw they should reconcile and you can keep the functionality you mentioned. The classes you’d want to look at are Ellipsoid, OpenEllipsoid, and ClosedEllipsoid. Open/Closed just determines if the points on the boundary of the shapes are considered “contained” by the shape. I’ll look into your code more in depth when I’m in tomorrow and let you know my thoughts!
Curtis Rueden
@ctrueden
Nov 28 2017 20:04
Thanks @awalter17! @kapoorlab We can also discuss further with @tpietzsch at the upcoming hackathon. (Even if you personally won't be there, we can look at it—and/or you could video call in one day to go over things.)
Varun Kapoor
@kapoorlab
Nov 28 2017 20:29
@ctrueden @awalter17 @tpietzsch Sounds great, I will be there at Neubias too and we can surely go over it before as well. :+1: