Custom Filter

[emencz]

How do you use it?

[Colin Walls]

I believe its quite an obscure feature. You need to get a book like PS Bible to get any details.

[Phosphor]

Whew....for the third time this week:

<http://www.creativemac.com/2002/02_feb/tutorials/cmfilterfactorycomplete.htm>

[Ol' Whozit]

Actually, Colin, if you check out the Help files in PS regarding Filter/Other/Custom you will discover that Phos' link explains the convolution properties discussed in the help files to a much better degree, and DOES fit what emencz is looking for, WITHOUT the extra expenditure for an additional book.

[emencz]

Thanks for the link, Phodphor. Sorry about the 3 times this past week thing.

And yes, I did mean filter>other>custom. Does anybody use it? I've seen some tuts regarding this feature, but all they do is tell me what values to enter.

[Phosphor]

No worries, E-Mendch! ;o)

I don't know how to use it either, and quite frankly, I don't care. I'm just happy that there are number-crunchers out there that are able to put a pretty face on image manipulating algorithms so I can look like a whiz, doing what I do.

Look around for Gernot Hoffman's name. He's pretty good at that stuff, and I'm sure he'd be happy to spell it all out for you.

[Colin Walls]

I stand corrected.

[Gernot Hoffmann]

Thanks for the compliment.
A filter works as follows:
The numbers in the box are weighting factors.
The center pixel is replaced by the weighted sum of all
pixels in the box (empty=zero weight)

E.g. a blurring filter, sort of primitive Gaussian:
1 2 1
2 7 2
1 2 1

The problem in PhS is the scaling - doesn´t work automatically:
Scale by the sum of all factors: 7+4*2+4*1 =19
Offset 0

The most "stupid" blur filter is the box filter:
1 1 1
1 1 1
1 1 1
Scale=9

This is a highpass, the sum of all factors is 0. A bad highpass.
-1 -1 -1
-1 +8 -1
-1 -1 -1

Scale=1 (not understandable), offset=128.

Sharpening filters are a little more complex.

A rather simple illustrated explanation is here.
<http://www.fho-emden.de/~hoffmann/gauss25092001.pdf>

Best regards --Gernot