Dear Vasily,
Segmentation of such structures may be quite difficult, especially if
the objects are touching.
You may also try the "iterative thresholding" plugin that segments
objects based on their size.
For the merging method, that is quite a interesting method also, you may
want to have the list of regions touching a specific region using
"closest object".
You can also try different settings, one with oversegmentation, one with
undersegmentation, you can see in a big object from undersegmentation,
what is the list of object in undersegmentation. If you send me an
image, I can have a look.
Best,
Thomas
On 25/02/2016 17:50, Василий Попков wrote:
> Good time of a day, friends.
> I need an advice on 3d segmentation.
> I need to segment mitochondria (mostly elongated "worms", sometimes
> rounded). I have found, that all plugins I have used (mostly I worked with
> spot segmentation) are either under segmenting or oversegmenting, while
> almost every plugin select all "needed volume".
> I think the easiest way to correct it is going from over segmented variant
> and merging
> all, what needs to be merged. The problem is that I havent found easy and
> fast way to do it. There are from 300 to 1500 segments in the cell. I can
> merge them in 3d manager, but it is not very fast: I need to check via 3d
> viewer which objects need to be merged, like "1, 126, 1354, 1478" find them
> in list and merge. I cant, for example "shift click" on objects in 3d
> viewer and say "merge", which will be like 10 times faster. This is
> possible way, but awfully slow for such numbers of objects.
>
> So. I am sure I am not the first one to face this problem. If anyone can
> direct me to plugins, which can suit my needs or give some advices what I
> am doing wrong - it will be very helpful.
>
> Thank you a lot for your time,
> Vasily
>
--
/***************************************************************/
Thomas Boudier, Associate Professor, UPMC,
Université Pierre et Marie Curie, Paris, France.
BioInformatics Institute (BII)/IPAL, Singapore.
/**************************************************************/
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html
Locked
