Strange Threshold on an image

[lechristophe]

Dear all,

I've a problem with the Auto Threshold setting with images that have
quite a lot of saturated pixels (I know it is bad imaging, but it is
not one of my images :) The image is available here :

http://ftp.espci.fr:21/incoming/christo/Strange_Threshold.tif

It is a 8-bit 1024x1024 image. If I Auto Threshold it, it sets it to
0-99, i.e. it detects what is OUTSIDE of the object (if I binarize,
I've the outside of the cell at 255 and the inside at 0). To be
consistent with non-saturated images, the threshold should be 99-255 I
guess. The problem doesn't exist with non saturated images : if you
divide this image by 3, the Auto Threshold detects the inside of the
cell, not the outside (Auto Threshold sets to 41-255).

Is it a rational consequence of the Auto Threshold algorithm or is it
a bug ? I've modified my macro that batch treats images to detect
these cases (detect and correct if the lower threshold is set to
zero), but I wonder if this behavior has a rational explanation.

Thanks for your help,

Christophe

[lechristophe]

Sorry for the bad link, the image is here :

http://ftp.espci.fr/incoming/christo/Strange_Threshold.tif

On Thu, May 29, 2008 at 10:39 AM, Christophe Leterrier
<[hidden email]> wrote:

> Dear all,
>
> I've a problem with the Auto Threshold setting with images that have
> quite a lot of saturated pixels (I know it is bad imaging, but it is
> not one of my images :) The image is available here :
>
> http://ftp.espci.fr:21/incoming/christo/Strange_Threshold.tif
>
> It is a 8-bit 1024x1024 image. If I Auto Threshold it, it sets it to
> 0-99, i.e. it detects what is OUTSIDE of the object (if I binarize,
> I've the outside of the cell at 255 and the inside at 0). To be
> consistent with non-saturated images, the threshold should be 99-255 I
> guess. The problem doesn't exist with non saturated images : if you
> divide this image by 3, the Auto Threshold detects the inside of the
> cell, not the outside (Auto Threshold sets to 41-255).
>
> Is it a rational consequence of the Auto Threshold algorithm or is it
> a bug ? I've modified my macro that batch treats images to detect
> these cases (detect and correct if the lower threshold is set to
> zero), but I wonder if this behavior has a rational explanation.
>
> Thanks for your help,
>
> Christophe
>