Page 3 of 3

Posted: Wed Jan 26, 2011 10:35 am
by leeharper_admin
Definitely the horizontal curve for CM4.


That's wonderful news Mike; I'm sure that we're all going to like that ;D

Cheers,
Lee.

Posted: Wed Jan 26, 2011 6:44 pm
by sjordan93436
The horizontal line sounds great.  But, doesn't the line have to (should?)  wrap around.  The value of 359 should be next to the value of 0.  The right intercept should equal the left intercept.  They don't have to, but isn't that best?

If that is true, can you do the same thing with your x/y curves?  0,0 to 359,359.  If you move the 0 intercept (left edge) away from 0,0 to say 10,0, shouldn't 359,359 move also?  To say 10,359?  I could do some photoshop to illustrate.

I would think the horizontal line would be best, but what if the alternate was far, far easier to program and almost as good.

Excuse my ramblings.

Posted: Wed Jan 26, 2011 8:22 pm
by leeharper_admin
Hi Steve,

The horizontal lines for the HSB curves will wrap around; Greg mentioned in an earlier post that one of the disadvantages of the current Hue curve is that if you try to change Reds on the Hue curve, you need to move both ends of the curve; my understanding of Mike's intent (from his replies) is that the ends of the horizontal curve line will be linked - so this will no longer be a problem.

As for your question about x/y curves - I'm not sure quite what you're asking? I think that an illustration would help us all...

Cheers ;)
Lee.

Posted: Wed Jan 26, 2011 8:31 pm
by -default
Yes, the ends would be horizontal and would be linked together.  Essentially the vertical axis of the graph would now represent not the hue values themselves, but an offset from the original hue value, with the initial "curve" being set to zero at all points.

Steve brings up an interesting poing.  I actually did think about keeping a more conventional x,y curve, but the problem is there is no simple way to link the ends.  For example, moving hue 0 up to hue 10 would require a corresponding upward change to the right side of the curve, which is already in the upper right corner.  The only way I could think of to address this involved either breaking the curve line into segments, or tiling the curve area with replicas of the curve area.  If Steve, or anyone else, has seen a simpler solution I'd be very interested in a mock-up.

Posted: Thu Jan 27, 2011 12:00 am
by sjordan93436
My point was poorly thought out.  Lee, you could not understand because the thoughts were bad.  Ignore it, but ....

Let us say we have a grid x horizontal, y vertical. 

X=  0 to 359  hue (original hue)  Every original hue needs a corresponding value.

y from 300-359, 0-359, 0 - 60  IOW, standard Y but with head room and basement. 

Starting with a straight line from 0.0 to 359, 359 with points at each end.

Say you grab a 350 reddish tone (magenta side).  Move it up _past 359 to say 30.  the curve would go into the headroom and back to 359,359.

Then you can move 359 to say 45 and the 0 goes to 46. 

You can have the rainbow on the x axis and y axis.  You can extend the headroom and basement 20,30, 60 whatever. 

Does this make sense?  I have NO idea if this is easier on Mike or better for users.  I was thinking out of the box.

When I get home, I might try a mock up.  this could end up like the MC Escher illustration.....

Posted: Thu Jan 27, 2011 1:10 am
by -default
Ah, I understand.  Thank you.

The head and basement room provide some wiggle room for wrapping the two ends.  The nice thing about this is that the graph still looks about the same, and you can still wrap the two ends together.  Very nice.  Not sure if I like it better than just the horizontal line, but it is certainly practical and workable.

Posted: Thu Jan 27, 2011 1:23 am
by sjordan93436
Just happy to give you an option.  Good luck.