Photoshop Blur function

Hi,

Does anyone know how the Blur and Gaussian blur filters are implemented (I mean what is the mathematical calculation for each pixel). In particular what exactly is the radius in the Gaussian blur and how is it used in the calculation.
Or do you know where can I find this information?

Many thanks,
Son

Hi,

Does anyone know how the Blur and Gaussian blur filters are implemented (I mean what is the mathematical calculation for each pixel). In particular what exactly is the radius in the Gaussian blur and how is it used in the calculation.
Or do you know where can I find this information?

Many thanks,
Son

The "radius" in Gaussian blur is equal to the sigma of the gaussian kernel.

But the mathematical details are not documented, for obvious reasons.

Dear Chris,

MAny thanks for your reply.
I need to know the exact calculation since I'm using the images in a scientific work... do you know who can I contact to find out the exact calculation? (I tried to reproduce the Baussian blur using the radius as the sigma and reached different results..)

Thanks again
Son

Any basic book on image processing will give you the information you need.

""" писал(а):
"""

Dear Chris,

MAny thanks for your reply.
I need to know the exact calculation since I'm using the images in a scientific work... do you know who can I contact to find out the exact calculation? (I tried to reproduce the Baussian blur using the radius as the sigma and reached different results..)

Thanks again
Son

Usually Radius is a Sigma*3
But I do not belive that Adobe implemented "true" Gaussian blur. If convolution kernel is critical for your applications (for example inverse problem solution etc)
I do not recommend use it.

Dmitry

------------------------------------------------------------ ----------- Software for digital image processing
http://www.imageskill.com

<http://en.wikipedia.org/wiki/Gaussian_blur>

Hmmm! What is interesting on the Wiki page is that the link to cee.hw.ac.uk (Heriot Watt University) seems to be bounced to Edinburgh Uni. OT but interesting.

Thanks Chris and Michael for your relpies. I was not asking how to calculate the Gaussian blur in general (this I know, and tried to implement myself and got results different from what I get with Photoshop). I was asking how to get information on how it is implemented in Photoshop. Anyone has an idea who I can contat to ask?

Thanks again,
Son

Look at your startup screen, then at the name of the posters in this thread...

Well, you can submit your resume to Adobe, and try to join the Photoshop team....

Ouch

Son,

In case you aren't getting the point Chris is trying to make here, you are asking for the recipe for Coca Cola. Nobody at Adobe is about to give up trade secret information which is what you are asking for - even if your purposes are purely benign and academic in nature.

You are out of luck.

Unless he has an acceptable (to Adobe) CV/resume, an aspiration for Adobe to appear on the wageslip and with suitably qualified trust and integrity?

Isn't there a website somewhere that tried to compare the algorithms used in commercial image processing packages?

I seem to recall seeing it once or twice with suitable example images alongside.

Son,

'I tried to reproduce the Gaussian blur using the radius as the sigma and reached different results..'

The reason is, that the filter box can have any width
by pixels, 3,5,7,..
There is no unique assignment of the Gaussian bell to
the pixels.
Just a general rule: larger sigma - larger box.
The filter weight factors are anyway only samples of
the Gaussian bell.
We have always approximations. Nothing bad, but always
slightly different.

<http://www.fho-emden.de/~hoffmann/gauss25092001.pdf>

Best regards --Gernot Hoffmann

Gernot - there is a unique assigment, because we do fractional sigma values and integer distances, and the radius in the UI is exactly equal to sigma for the gausssian distribution used to create the filter.

<SteveV>"I'm almost certain that means stuff"</SteveV>

;) XD

(and that was the simplified version)