Possible bug in Edit->Selection->Create Selection

[Neil Fazel]

I have a thresholded grayscale image on which I used Edit->Selection->Create Selection; the selection which was created is not closed, i.e. it does not fully contain the thresholded region, as I think it should. Also, when I use getSelectionCoordinates() on it, it returns only 4 points, whereas the selection has many more vertices.

To reproduce, please

1) import the .txt image into ImageJ,
2) threshold it (using 0.644 and 77 for the lower & upper threshold levels, resp.).
3) use Edit->Selection->Create Selection to create a selection.

You will see that the selection skips thresholded regions at the bottom of the image.

I tried 3 times to attach the relevant files (.txt, .roi, .tiff) necessary to reproduce the problem, but each time ImageJ server rejected them (it doesn't like .roi or .zip files). Who should I email them to?

Vielen Dank!
Neil

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Michael Schmid]

Hi Neil,

when working with 16-bit or 32-bit images, the 'red' display of the thresholded area is only an approximation.  It simply updates the color lookup table, so there are only 156 distinct values of the threshold where it is correct.  Maybe that explains your problem?

--

For showing a file to others on the mailing list, it might be the easiest to upload it to a web server (your own one, if you have one, or some dropbox-like service).  I guess that the mailing list only accepts rather small attachments.

Michael
________________________________________________________________
On Mar 6, 2014, at 19:21, Neil Fazel wrote:

> I have a thresholded grayscale image on which I used Edit->Selection->Create Selection; the selection which was created is not closed, i.e. it does not fully contain the thresholded region, as I think it should. Also, when I use getSelectionCoordinates() on it, it returns only 4 points, whereas the selection has many more vertices.
>
> To reproduce, please
>
> 1) import the .txt image into ImageJ,
> 2) threshold it (using 0.644 and 77 for the lower & upper threshold levels, resp.).
> 3) use Edit->Selection->Create Selection to create a selection.
>
> You will see that the selection skips thresholded regions at the bottom of the image.
>
> I tried 3 times to attach the relevant files (.txt, .roi, .tiff) necessary to reproduce the problem, but each time ImageJ server rejected them (it doesn't like .roi or .zip files). Who should I email them to?
>
> Vielen Dank!
> Neil

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Neil Fazel]

Hi Michael,

   Thanks for the tip. To look into it, I thresholded differently, i.e. between 0 and and 0.01 (with Dark Background option selected). There are contiguous regions of the image which are all 0's (e.g. the thin vertical lines on the right-hand side of the attached .tiff image); the selection drawn for them is a rectangle open on one side (bottom, visible when magnified). I would have expected a region of uniformly 0 pixel values to be fully encompassed by the selection since, it seems to me, there are no thresholding approximations involved.

Thanks,
Neil


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Neil Fazel]

I figured out that the selection is in fact closed; the lower boundary falls outside the image and is not normally visible. When I scaled the selection down by 1% (Edit->Selection->Scale), the full selection boundary became visible. (Image attached.)

The question remains as to why getSelectionCoordinates() returns only 4 coordinates. Perhap it can't handle composite selections. If so, would switching to ImageJ2 help?

Thanks,
Neil


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html