Threshold rules

[Johannes Breu]

Hello.

As it is a general problem that background disturbes the measurement of
signals (e.g. fluorescent signals) and that a rough thresholding reduces the
signal I wish to ask for some rules according to this problem. I want a
better signal/noise ratio.

My results are completely unsatisfactory. Judging by eye I see a specific
signal but if I measure the signals I get in trouble. I am irritated how to
find the best threshold parameter. As an alternative I thought to substract
intensities before thresholding. Another possibility: Is there a way to
treat all images the same way that the signa/noise ration gets better before
I start with thresholding?

Thanks.

[Gabriel Lapointe]

Depending on what you are looking at a simple mean or median filter
might be enough  to  improve your s/n ratio. For  3d stack I find the 3d
hybrid median filter to be more than suitable for my confocal stack.
Khalman algorythm might be another option.

Gabriel

J Breu wrote:

> Hello.
>
> As it is a general problem that background disturbes the measurement of
> signals (e.g. fluorescent signals) and that a rough thresholding reduces the
> signal I wish to ask for some rules according to this problem. I want a
> better signal/noise ratio.
>
> My results are completely unsatisfactory. Judging by eye I see a specific
> signal but if I measure the signals I get in trouble. I am irritated how to
> find the best threshold parameter. As an alternative I thought to substract
> intensities before thresholding. Another possibility: Is there a way to
> treat all images the same way that the signa/noise ration gets better before
> I start with thresholding?
>
> Thanks.
>
>  

[Sudipta Mukhopadhyay]

Hi

It is a problem of Image Restoration. There are rules in the book, but they
do not work always to your satisfaction. Looking at the images, it will be
easier to suggest the way.

With warm regards

Sudipta Mukhopadhyay

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of J Breu
Sent: Wednesday, October 22, 2008 8:31 PM
To: [hidden email]
Subject: Threshold rules

Hello.

As it is a general problem that background disturbes the measurement of
signals (e.g. fluorescent signals) and that a rough thresholding reduces the
signal I wish to ask for some rules according to this problem. I want a
better signal/noise ratio.

My results are completely unsatisfactory. Judging by eye I see a specific
signal but if I measure the signals I get in trouble. I am irritated how to
find the best threshold parameter. As an alternative I thought to substract
intensities before thresholding. Another possibility: Is there a way to
treat all images the same way that the signa/noise ration gets better before
I start with thresholding?

Thanks.