Posted by
Justin McGrath on
Jan 15, 2007; 4:38pm
URL: http://imagej.273.s1.nabble.com/use-of-imageJ-for-herbivory-assessment-tp3700587p3700591.html
Hi Jonathon,
Are you thresholding the images first? The wand tool will only select
a group of pixels that have the same value. If your pictures look
good and have even illumination and good contrast with the background,
convert them to greyscale then use Image->Adjust->Threshold. If
they're already binary, try clicking slightly outside of the leaf
instead of in the interior. I find that it works better.
If you're finding it hard to threshold the images because of uneven
illumination, I have a plugin that uses something like NDVI to create
an index of greeness that may make it easier.
You probably already know this, but just in case. When you use the
measure command, it doesn't include interior holes, so you'll have to
measure each hole and subtract from the total. I don't know of a way
to make it not include interior holes (the analyze particles tool can
do it though), but it would be a nice addition to the measure command.
Justin
On 1/15/07, Jonathan Rosenthal <
[hidden email]> wrote:
> Hello all,
> I'm attempting to use ImageJ to analyze scanned images of leaves in
> order to assess percentage of leaf area lost to herbivory. In so
> doing, I've encountered two problems already (and I've just started):
> 1) The "wand" tool, selected from the toolbar, will, at most outline
> only one hole per leaf, will not outline the leaf perimeter, and, more
> generally, is often completely non-responsive; and 2) in trying to use
> a mouse or trackpoint to control the "freehand" tool from the toolbar
> to outline the leaf perimeter (or reconstruct an eaten leaf perimeter),
> my hand clearly is not steady enough to closely follow the path that I
> desire. I would greatly appreciate it if anyone would suggest
> either alternatives to these 2 tools, or ways that I can use them more
> effectively. Also, if you know of a standard protocol for herbivory
> assessment that's compatible with the current version of ImageJ, that
> would probably solve all of my problems.
> Thanks very much,
> Jonathan Rosenthal
>