Most mysterious filters

[Doug Nelson]

What are the Photoshop builtin filters that you find the hardest to understand?

[DannyRaphael]

Originally Posted by Doug Nelson

What are the Photoshop builtin filters that you find the hardest to understand?

Distort > Wave from the perspective it's nearly impossible (for me, anyway) to predict the outcome even though I pretty much understand the controls.

[gina]

For me that would be the render lighting filter, though I think it's a filter that can be tremendously useful.
Oh well, perhaps you should have really a lot of experience in lighting problems to fully appreciate the strength of it

gina

[byRo]

High-pass filter. I use it and I don't understand it.
(Which is a situation that makes me uneasy)
Good thing the PS7.0 Help clears it up:

High Pass (Photoshop)

Retains edge details in the specified radius where sharp color transitions occur and suppresses the rest of the image. (A radius of 0.1 pixel keeps only edge pixels.) The filter removes low-frequency detail in an image and has an effect opposite to that of the Gaussian Blur filter.

(That 0.1 is a typo, isn't it?)

Now if we were to include Image>Adjustments - I could write a whole speech against the "Color Balance".
Ro

[JustChecking]

Originally Posted by byRo

High-pass filter. I use it and I don't understand it.
(Which is a situation that makes me uneasy)
Good thing the PS7.0 Help clears it up:

(That 0.1 is a typo, isn't it?)

Now if we were to include Image>Adjustments - I could write a whole speech against the "Color Balance".
Ro

what's the problem? it's just like in functional analysis [all right, switching the nerd mode off ]... it just takes a "circle" of given radius around the point, and if the pic doesn't "change enough" there (like if there are no sharp edges etc.), it puts gray colour there... otherwise it puts a colour "corresponding" to the amount of the change there... that's the whole miracle - it's similar to the way you'd trace edges outta that pic - just where there ar sharp changes in colour/intensity(=edges), you draw...
and the 0.1 isn't typo - it's just the minimal value you can choose, i don't think it has any correspondance to the algorithm they use... just a number on the scale...



oh, and to answer the question... all LOL i haven't even tried them all out yet

[byRo]

Originally Posted by JustChecking

..otherwise it puts a colour "corresponding" to the amount of the change there

OK, when the radius is pretty small - then you can think of it as some sort of first derivative function where the result will represent the amount of change. But what gets me is why, for larger radii, the output image reverts to the original?
I think that the "descriptive" name is the best way to understand it because it relates well to a high-pass audio filter - if you lower the cut-off frequency more of the original gets through.

Originally Posted by JustChecking

...and the 0.1 isn't typo - it's just the minimal value you can choose

OK, true that it is the minimal value, but when you apply it to an image you get absolutely nothing (just mid-grey). Using a value of 1.0 you'll get something like the result of a simple Laplace filter - all edges and noise.
Getting back to Doug's question:
- if we have to resort to "nerd" talk;
- rely on analogies from other areas;
- the help text gives us something that just doesn't work.
then it is pretty hard to understand.

for JustChecking: Seems that you work in this general area, so let me ask: There's a lot of "rocket science" image processing going on (Fast-fourier, C/C++ libraries etc..) - is there a way that we "mere mortals" can use some of this stuff?
Rô

[JustChecking]

Originally Posted by byRo

There's a lot of "rocket science" image processing going on (Fast-fourier, C/C++ libraries etc..) - is there a way that we "mere mortals" can use some of this stuff?

well, there's lots of this stuff on the net, but most of it involves some programming to be done... (for a start you can check out http://www.oonumerics.org/oon/ , but i'm afraid many links will be dead by now)... as for programs i can't help you too much - on the faculty we use mostly commercial libraries for matlab or mathematica or libraries we write ourselves...
otherwise, if you can get your hands on it (eg. in your library or so; buying it isn't worth the money), try the IEEE Transactions on Image Processing - everything that can be said about image processing from the theoretical point of view is there... unfortunately, often it's pretty hard to find it there, and unless the algorithm is proposed, also implementation is typically tough... and of course CiteSeer - http://citeseer.ist.psu.edu/

if you're interested in something in particular, i can try finding out something about it...