Subtract Background algorithm

[Pavan-7]

Hello all

I am having some grayscale images that I am trying to make the intensity
gradient uniform accross each image and I am able to do this using
subtract background with rolling ball radius at 50 pixels, light
background checked and using sliding parabloid gives a little better
result but even without that it looks fine. I would like to know the
algorithm behind this or any good book for reference highlighting this
algorithm in detail. Also can someone explain me how this subtract
background is being implemented in ImageJ with the above options selected??

thanks

[Gabriel Landini]

On Friday 08 Jan 2010  14:41:52 you wrote:
> I am having some grayscale images that I am trying to make the intensity
> gradient uniform accross each image and I am able to do this using
> subtract background with rolling ball radius at 50 pixels, light
> background checked and using sliding parabloid gives a little better
> result but even without that it looks fine. I would like to know the
> algorithm behind this or any good book for reference highlighting this
> algorithm in detail. Also can someone explain me how this subtract
> background is being implemented in ImageJ with the above options selected??

You could start by looking in the documentation at Wayne's site, where it is
specified exactly what algorithm is used, who wrote it and where it is
published.

G.

[Johan Henriksson-2]

Pavan wrote:

> Hello all
>
> I am having some grayscale images that I am trying to make the intensity
> gradient uniform accross each image and I am able to do this using
> subtract background with rolling ball radius at 50 pixels, light
> background checked and using sliding parabloid gives a little better
> result but even without that it looks fine. I would like to know the
> algorithm behind this or any good book for reference highlighting this
> algorithm in detail. Also can someone explain me how this subtract
> background is being implemented in ImageJ with the above options selected??
>
> thanks
>  

algorithm is trivial, theory less so if you are interested. I think most
people apply it without thinking too much about the details. this is in
effect a high-pass linear filter, you can find (the one-dimensional
analogue) described in books on digital signal processing and control
theory. you can calculate optimal filters if you know enough about your
signal but it's usually not the case.

/Johan