Wow - blows my mind

Jacob Rus's new curve-based technique, with examples
-default
Posts: 1916
Joined: Thu Mar 26, 2015 1:53 am

Postby -default » Thu Sep 04, 2008 5:00 am

I've done some thinking about this, and the procedure seems to be very similar to using the Lightness channel as a mask, then curving the L, a, or b channels.

I haven't had a chance to veryify this yet.

mikemeister_admin
Posts: 4927
Joined: Fri Sep 20, 2013 8:29 pm

Postby mikemeister_admin » Thu Sep 04, 2008 9:26 am

You are probably right Mike, but I still do not understand why because....

The first image, below, is the Street scene showing the 3 Lab channels - the top row being the original and the bottom when the L channel has been placed over the a&b ones.  It is this bottom image that we are curving, once it has been normalised to neutral.

The next screen shot shows me curving arbitary multiple neutral points in LAB (note the screenshot has altered the profile and upset the neutral values a bit), but this demostrates the idea.

The last 2 are the curves (!) for the a and b channels to do the above.

So this has allowed me to set neutrals for different tonal values in Lab - something that I thought was impossible!

By choosing different Overlay type blends, then the amplification maths would be different - and even worse, there is no reason why I have to use the same 'mask' channel for each of the a & b channels - any of the 11 channels could be used! 

The variations are endless and I think allows me to select the best 'mask' for each channel to solve a difficult image.  It is this flexibility, that blows my mind as well as the rather inituitive curving of colour by tonal value (when the L channel is used).

For fun I tried curving the Jag car using this idea and used the a channel as the 'mask' - this allowed me to curve for any colour as the 'mask' allowed me to restricted my curving to just the car.

The downside is that, I believe, we are getting into impossible Lab colours and so twiddling one channel can upset the other as PS struggles to convert back to the rgb display space - this is all a bit too much for my tiny mind!

However, is it significant that the Colour Theory Forum has completely ingored this suggestion? 
Either it is because this is just another way of masking that does not amount to much, or it is a case of 'it wasnt invented here' reaction.

I'm looking forward to you putting me out of my misery and explaining what is going on.

mikemeister_admin
Posts: 4927
Joined: Fri Sep 20, 2013 8:29 pm

Postby mikemeister_admin » Thu Sep 04, 2008 9:35 am

Sorry, I attached the wrong image for the Lab channels
one needs to view them in grey scale not colour to see what one is curving

-default
Posts: 1916
Joined: Thu Mar 26, 2015 1:53 am

Postby -default » Thu Sep 04, 2008 9:38 am

I still haven't gotten to play with this yet, but I do think you are right - this may well be something fundamentally new.  Very exciting.

-default
Posts: 1916
Joined: Thu Mar 26, 2015 1:53 am

Postby -default » Thu Sep 04, 2008 10:15 am

I wouldn't read much into the silence of the CT folks.  Dan may be away, and he always takes his time before responding to something deep and new, which I now believe this to be.

At the very least, it's a mind bender, as you say, because it packages several separate concepts into one little bump on a curve.  This kind of revelation does not happen more than once every few years.

BTW, I love those hue clocks, zog.

mikemeister_admin
Posts: 4927
Joined: Fri Sep 20, 2013 8:29 pm

Postby mikemeister_admin » Thu Sep 04, 2008 10:44 am

re the clocks Mike,

The trouble is they are screen based and so not that accurate, especially in the greens and not a patch on CM, but I find them useful sometimes, particularly when I cant use CM - as in this technique.

I rather like this mode for fun Lab corrections where the circles show the lab values!

mikemeister_admin
Posts: 4927
Joined: Fri Sep 20, 2013 8:29 pm

Postby mikemeister_admin » Thu Sep 04, 2008 6:52 pm

Wow indeed. Yes, a radical approach -- I must admit it sounded senseless at first, when I just read it. But I now have, I believe, refined this method slightly, by using Soft Light instead of Linear Light as the blend mode -- that gives much more subtle control.

mikemeister_admin
Posts: 4927
Joined: Fri Sep 20, 2013 8:29 pm

Postby mikemeister_admin » Sat Sep 06, 2008 1:19 pm

Okay, an action.

And some screenshots in sequence:
jacob%27s_ladder.gif

mikemeister_admin
Posts: 4927
Joined: Fri Sep 20, 2013 8:29 pm

Postby mikemeister_admin » Sat Sep 06, 2008 7:51 pm

Interesting and different from your initial posting (assuming I read it correctly).

I do not like the Lab channels being shown back to front in our action!  b-a-l
and also it would be a heap better to just have one adjustment curve layer, rather than having to switch between them, which is extremely slow.

But for me the beauty of your simple first approach has gone - adjusting curving using the same seed mask.  This is extremely intuitive as any point on the image is at exactly the same point on all 3 curves - not as powerful as your action, but great for relatively simple adjustments.

Am I also correct in thinking that by effectively doing an auto-levels on the channels, there are now big gaps in the histogram?  I had thought of doing this and came to the conclusion that a bit of blurring would be in order to smooth things out again.  Perhaps I am missing something here.

There certainly is lots of room to experiment with your initial idea - but for simple people, like me, I need to pre-visualise what will happen when I curve - your action does not give me this.


mikemeister_admin
Posts: 4927
Joined: Fri Sep 20, 2013 8:29 pm

Postby mikemeister_admin » Sat Sep 06, 2008 9:46 pm


Interesting and different from your initial posting (assuming I read it correctly).

No, it's identical.
Am I also correct in thinking that by effectively doing an auto-levels on the channels, there are now big gaps in the histogram?  I had thought of doing this and came to the conclusion that a bit of blurring would be in order to smooth things out again.  Perhaps I am missing something here.

All this does is zooms the horizontal axis of the editing curve, clipping it to the region where moving the curve will have any effect.  Because this layer is only used for adjustment, beginning by default with no effect, it doesn't really matter very much that there are gaps in the histogram.  I always use 16 bit/channel mode to avoid rounding errors here, but I'm not sure it makes too much difference.
There certainly is lots of room to experiment with your initial idea - but for simple people, like me, I need to pre-visualise what will happen when I curve - your action does not give me this.

Why can't you previsualize?  If you click around in the image while the curve is up, you can see which parts of the image correspond to which part of the curve, and then by knowing what the image looks like, what its numbers are, etc., you can decide what you want to do with each part of it.

I think one thing which might make an improvement here would be to figure out a way to add three (step?) gradients at the bottom of the image to make it easier to see precisely what's being done to specific colors.
This is extremely intuitive as any point on the image is at exactly the same point on all 3 curves - not as powerful as your action, but great for relatively simple adjustments.

This is still the case.  If you want, just chop the a* and b* layers out of the action, or delete them once it is complete.
and also it would be a heap better to just have one adjustment curve layer, rather than having to switch between them, which is extremely slow.

If one adjustment curve layer could fit 9 curves on it, I'd totally jump on that.
I do not like the Lab channels being shown back to front in our action!  b-a-l

Okay, I've uploaded a revised action, with the order reversed.

Also, an identical action with explanatory stops inserted.


Return to “Jacob's Ladder”

Who is online

Users browsing this forum: No registered users and 7 guests