Filter Forge

[Cameraken]

Hi Rô.

Invert the Original
Of Course, Why didn’t I think of that?

I was thinking the wrong way with this; I was hoping to build a 5x5 filter that I could re-use. But it seems that this will need re-working for each kernel. I will stay with 3 x 3 in future.
Offset Percentage. I understand that now (I think).
Linear Light is on Vladimir’s too do list

Thank you for the example. That makes everything a lot clearer.

I have managed to make a rectangular frame but it’s a bit messy I guess we would be better waiting for ‘bombers’


Ken.

[Stroker]

I think I found a bug. But it could be a feature.
Levels won't invert if I swap the Black Point and White Point sliders.
Annoying, but survivable.

[byRo]

Ken, nice to see you got some filters admitted.
Passing on some advice from Vladimir, you could make your presets a little more interesting. For example the basic Tri-tone filter works fine and well-behaved, now just add in some extreme stuff in the presets to "sell" it. Maybe three different colours.

Did you try any alternative to the three impulse curves?
These make step changes between each colour. I would have expected some sort of fading.

Hi, Stroker. I'm pretty sure that it's feature, there are some other components that do that too.
Seems that it is FF policy to (over)protect us from doing something "wrong".
Agreed that it rules out some fun stuff. We'll just have to be more inventive.

Rô

[Steve Conway]

Ok, Ro...I'm giving it a shot. And I appreciate the feedback.

Now you're really gonna see some bad filters.

Steve

Originally Posted by byRo

Steve, I saw this too - but I also saw somewhere that you can turn it off.

Doesn't worry me too much though.
Vladimir and the guys are serious folk and they are only interested in feedback about their filter.
(or they fooled me real good!)

Rô

[Steve Conway]

That sets me straight. Glad you posted this, I appreciate it. Sometimes you get paranoid regarding software downloads that are free.

Plan to load it shortly and probably make some filters so bad they wouldn't want to know anyway. 8-)

Side note: if anyone comes up with any painterly-type filters either by accident or on purpose, I would appreciate some pointers on these.

Again thanks.

Steve

Originally Posted by Cameraken

Hi Steve.

Here is the information about the info that FF use.
http://www.filterforge.com/more/usagestats.html

However it can be switched off in the preferences if you wish. There is also a message which asks if it’s OK to send the info before it is sent.

FF is worth trying. Some wonderful effects can be achieved. Just look what Vladimir had to say about Byro’s filter.



Ken

[byRo]

Originally Posted by Steve Conway

Plan to load it shortly and probably make some filters so bad they wouldn't want to know anyway. 8-)

If you wish, post them here first, Ken, Stroker, Craig* and I could give you some pointers.

Rô

*hope I didn't miss anyone.

[Kraellin]

If you wish, post them here first, Ken, Stroker, Craig* and I could give you some pointers.

just bear in mind that Ro, Ken and Stroker know far more than i do when it comes to the beast behind the beauty. i'm just here for the pretty pictures

craig

[Stroker]

It's a fun beast to play with. Just know that my main interest is in finding bugs. But if something does tweak my noodle, I'll speak up.

For example, you can use Offset to do what Displace does.

[Steve Conway]

Thanks for the offer of help guys.

Just a thought, after playing with it for a few hours. One of you beta testers need to tell them that as in so many image manip programs the work windows are too small. In this case too many and too small. And no way to see your work on the original until you apply it.

If you have Gertrudis. This is the kind of work space these programs need.

Not too enthused at the moment, but may take another look as time permits.

Steve

[byRo]

Steve, there is an active forum at their site.
There many people, as you, have criticised the usage of the window's "real estate".
The next version should be a little better.

Rô

[Steve Conway]

Ro, I'll go by and take a closer look. Did notice it was there, but have not gone into it for info and stuff.

Appreciate the tip.

Steve

Originally Posted by byRo

Steve, there is an active forum at their site.
There many people, as you, have criticised the usage of the window's "real estate".
The next version should be a little better.

Rô

[Cameraken]

Is it me or is there a bug?

Mike Blackney built a 5 pointed star but the code was too slow so I tried to speed it up however I can’t get the ‘maths’ to work.

The angles should be 72, 144, 216, 288 and 360 degrees for a 5 pointed star.

When blending two gradients the slider goes from 0 to 100 which equates to 0 to 90 degrees. So the slider must be set at 40 to equal 36 degrees and 80 for 72 degrees but when I use these settings the star is not correct. I can easily correct this but I would rather find what is wrong.

Clicking Linear or Smooth Gradients does not help although linear seems the best. Is my logic wrong or is there a bug?

Ken.

[byRo]

Originally Posted by Cameraken

When blending two gradients the slider goes from 0 to 100 which equates to 0 to 90 degrees. So the slider must be set at 40 to equal 36 degrees and 80 for 72 degrees but when I use these settings the star is not correct. I can easily correct this but I would rather find what is wrong.

Sorry, Ken. It's a bit more complicated than this. When you blend the orthogonal ( ) gradients @ 40%, that means you get 40% of one and 60% of the other.
Remember your "trig" classes? This is a right-angled tringle with the vertical side of 40 and the horizontal side of 60, so the opposite angle is arctan(40/60) - which is 33.69 degrees (and not 36).
Starting with 36 degrees and working back, you'll end up with a blend of 42.08%.
The other blend, instead of 80% is in fact 75.48%.

(the thresholds I just lined up by eye)

Rô

[Cameraken]

Hi Rô

Thank You. That solves the problem and answers my question perfectly.
I knew something was wrong, I just couldn’t see where. I have re-done my spreadsheet of angles and I have even made an angle calculator which I am attaching here. It’s made in VB and so requires VB run-time.

I have searched the net for info on gradients, offsets and curves but can’t find anything useful.
Now if I could only find such a good explanation of multiply, squared etc?

Re.
http://www.filterforge.com/forum/rea...=623#nav_start

I don’t think a Gamma Spreadsheet (or a zip file) will upload at FF. Maybe you could post them here. I would be interested in seeing them.
Is this any help?
http://www.rskey.org/gamma.htm

I am trying to keep up. What is “the square root trick using the Gamma curve”?

Thanks again.

Ken.

[byRo]

Originally Posted by Cameraken

I have searched the net for info on gradients, offsets and curves but can’t find anything useful.
Now if I could only find such a good explanation of multiply, squared etc?

OK, I hear you.

Originally Posted by Cameraken

.....“the square root trick using the Gamma curve”?

When making a more mathematically based filter, square and square root functions are very useful. There is an easy way to do the square, but the only way I know to do the square root is with the gamma function.

In the snippet attached the Gamma curve generates a sqr(c) (squared) output when we specify a gamma value of -30.10.
Feeding that into the gradients and adding (blend, normal, 50%) you get a gradient of sqr(x)+sqr(y). That's good, but for it to be linear we need the square root of that. Running it through the Gamma adjustment, value 30.10, we end up with a luvly linear radial gradient. r = sqrt(sqr(x) + sqr(y))

Problem is that the Gamma isn't exactly the square root - though only a nitpicker would worry about that. (but, seeing as I am one.........)

(Excel file attached)

Rô

[Cameraken]

Hi Rô

Thanks, I have studied and (I think) understand that. I guess as more of these snippets are written then the need to understand them fully will diminish. (We can just use them)
However, I like to have some idea of the ‘how it works’.

One (last?) question if I may, My Sepia Filter which (I agree) I should not have used an impulse curve to select shadows, midtones and highlights.
This filter requires some fading or smoothing in the transitions areas. (Like Stroker wrote for Craig’s filter in FM). What would be the best way to do this?


Attached is an example of the 5 point star angles in action. Happy 4th July America.

Ken.

[byRo]

Originally Posted by Cameraken

....I should not have used an impulse curve to select shadows, midtones and highlights.

No problem with the impulse curve.
What happens is that when you put in a value of 100% for the plateau, there's no room for manouver. Change the value to 0%, increase the width to 100% and use positions of -50, 0 and 50.
.......Nice smooth transitions.

The flag looks perfect to me! Great work!!

Rô

[Cameraken]

Thanks Rô, That’s much better

I was never happy with that filter, I will send in an update.

Ken.