> should consider a mask-based ROI type. Maybe the ImageJ2 developers
ROIs extend the same base RegionOfInterest interface. This includes
shape-based ROIs, but also thresholds and masks. That said, please note
> Hi VC,
>
> well, I don't have a computer system as powerful as yours, but I can
> imagine that an image of such a large size can need a long time for 'Create
> Selection' if it has many small particles.
>
> On my computer, 'Create Selection' with the 'Particles' sample image
> replicated 2x2 times (4000x2000 pxl) needs about 1.8 sec, but when
> replicated 4x4 times (8000x4000 pxl) it needs 34 sec. So there is some
> overhead that grows roughly with the square of the image size (or the
> number of particles). The progress indicator needs only a small portion of
> that time.
>
> Looking at the source code of ThresholdToSelection.java, I guess that the
> computing time is probably spent in the java.awt.geom.GeneralPath.append
> method: It is not based on pixels but on outlines, which may have double
> precision accuracy. When adding any new particle (outline) to it, it has
> to check thousands of other particles (polygons) whether they might touch
> or intersect. The computing time for it will be proportional to the square
> of the number of particles; for my case with ≈160000 particles that's about
> 10^10 tests. Even if these checks are done efficiently (checking the
> bounding rectangles only) this will take some time. In principle, this
> check is not necessary, but as far as I know one cannot avoid doing this
> check (also the new Java 1.6 Path2D methods do not help).
>
> ---
>
> I also tried to deselect all options in 'Set Measurements' and 'Analyze
> Particles' with adding to the ROI Manager. This takes less than 7 seconds
> for the 8000x4000 pxl image. However, combining all the 160000 ROIs takes
> ages (I finally 'killed' ImageJ because I did not want to wait for hours),
> so don't try this as a way out!
>
> ---
>
> The bottom line: if you have images with >100000 particles (or fewer in
> some cases), avoid selecting all particles at once. Rather use a mask and
> do operations with it. Alternatively, use 'Limit to threshold' in 'Set
> Measurements'.
>
> On the long term, if problems like this should be more common, one should
> consider a mask-based ROI type. Maybe the ImageJ2 developers read this...
>
>
> Michael
> ________________________________________________________________
> On Dec 9, 2013, at 16:17, VC wrote:
>
> > hi all,
> >
> > I'm creating a selection from auto threshold of a 541MB picture on late
> 2012
> > iMac, 32Gb (4x8) i7 processor (using Fiji, but it's the same in Ij2 or
> Ij1).
> > I increased memory to 20G in options menu and in plist.txt.
> >
> >
> > It takes over 20 minutes to do this. Is this normal??
> >
> > I added a print screen while the create selection is running.
> >
> > Thx in advance!
> >
> > <
http://imagej.1557.x6.nabble.com/file/n5005856/fiji.jpg>
> >
> >
> >
> > --
> > View this message in context:
>
http://imagej.1557.x6.nabble.com/is-this-normal-tp5005856.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>
Locked
