Elite Retouch wrote:
Sean....DerW...

Have you guys figured out how to apply this technique to an image in 16bit Lab Color Mode?

I wanted to test my 3 layer separation in that mode to see if the values held up more...but the same settings don't apply in the "Apply Image" menu.

Unfortunately Adobe hasn't implemented (or is restricted by the interface?) those blends in Add Image for LAB, so you have to do it 'old school', or manually do everything through Calculations and reconstruct your final image.  Use the Brightness/Contrast / Legacy / -50 Contrast + High Pass (filter) method outlined above to do it.  Do note that by default the B/C adjustment won't affect the a and b channels, so you have to individually select them and run the adjustment three times (once per channel), else you'll get a hyper-saturated result when you're done.  With quick testing, this method in LAB looks like it's accurate to ~3/32k per channel.

SRB Photo wrote:
Unfortunately Adobe hasn't implemented (or is restricted by the interface?) those blends in Add Image for LAB, so you have to do it 'old school', or manually do everything through Calculations and reconstruct your final image.  Use the Brightness/Contrast / Legacy / -50 Contrast + High Pass (filter) method outlined above to do it.  Do note that by default the B/C adjustment won't affect the a and b channels, so you have to individually select them and run the adjustment three times (once per channel), else you'll get a hyper-saturated result when you're done.  With quick testing, this method in LAB looks like it's accurate to ~3/32k per channel.

.........I think I'm gonna have pizza for dinner tonight.

Elite Retouch wrote:

.........I think I'm gonna have pizza for dinner tonight.

MMMMMMM.....pizza....

I normally have to read through Sean's posts a few times and do a bunch of Googling so I can understand.

MMMMMMM....pizza....

Photons 2 Pixels Images wrote:

MMMMMMM.....pizza....

I normally have to read through Sean's posts a few times and do a bunch of Googling so I can understand.

MMMMMMM....pizza....

I had no idea I was losing everyone so thoroughly - that's embarrassing .

SRB Photo wrote:

I had no idea I was losing everyone so thoroughly - that's embarrassing .

Not thoroughly. And it's only embarrassing to me. I should get this stuff a lot quicker than I do. Problem is, I haven't used this stuff at all for the past....I can't even count them....years and now that I'm into digital imaging which uses some of the theories I used to know I have trouble recalling them.

So, just keep doing what you're doing. I'm learning a lot from it. It's forcing me to read and research and that's a good thing. I needed a mental challenge at this point in my life. I'm certainly not complaining.

DerW wrote:
As LAB already separates the luminosity from the color, you would "just" have to separate the different frequencies of the L-channel.
So I would just duplicate the background layer two times, desaturate the bottom two layers (that works in LAB because it just fills the a- and b-channels with 50% gray, which is neutral) and blur the bottom's L-channel.
Switch to the middle layer, select it's L-channel and use "Image"-"Apply Image", once again, invert the L-channel from the bottom layer, use a scale of 2 and offset of 0. Set the blending mode of this layer to "Linear Light", switch to the topmost layer and set its blending mode to "Color".

I recorded an action to do exactly that: http://www.sendspace.com/file/675wdg

PS: Sean, how do you always calculate the accuracy so well? So far I always have to use the "Difference" mode and levels to exaggerate the whites, but that doesn't give me those precise numbers like yours.

Oh now that's interesting - you can use all the blend modes of Apply Image - if you go through each channel individually.  What a PITA!

As to the accuracy question, I'm doing much like you are, but I'm using the eyedropper and info palette set to 16bit readings to record the differences,  using tools like Curves / Levels / Threshold to find the maxima.  When I say something like "quick testing" it usually means 1-2 real world images; "rigorous" would be something deliberately designed to get as much error as possible.

W.G. Rowland wrote:
In another thread on noise removal apps I was asked about using separation techniques to handle noise (paraphrasing hugely).  That's gotten me wondering (but not enough to read through this whole thread again to find out)...

Has anyone played with per-channel separations to see about isolating certain details even further? 

After all certain channels do tend to hold on to more of certain things (blue tends to be noisies if I recall), so would using a mask based off of a separation of the blue channel allow you to hide more noise without killing as much other fine detail that happened to exist at the same radius?

Hope that made sense, if my mind hasn't wandered I'll be checking it out when I get home..

Now that's interesting. I hadn't thought of that approach. I have used separation to get rid of some noise and it works fine with a bandpass around the radius of the noise. I hadn't thought to mask using the blue channel, though, in order to target just the noise. I'll have to try that.

DerW wrote:
As LAB already separates the luminosity from the color, you would "just" have to separate the different frequencies of the L-channel.
So I would just duplicate the background layer two times, desaturate the bottom two layers (that works in LAB because it just fills the a- and b-channels with 50% gray, which is neutral) and blur the bottom's L-channel.
Switch to the middle layer, select it's L-channel and use "Image"-"Apply Image", once again, invert the L-channel from the bottom layer, use a scale of 2 and offset of 0. Set the blending mode of this layer to "Linear Light", switch to the topmost layer and set its blending mode to "Color".

THIS is what I've been talking about. This method produces no value shifts at all. EXCELLENT...but is there anyway to convert this information back over to RGB without losing the summation of any of the data? I tried merely switching it back, and the separated image looked a little something like what first time retouchers do to a photo and call it "Done".

Elite Retouch wrote:
I used to go into LAB mode to increase color vibrancy, and then come back to RGB to finish my editing. I was hoping that doing the 3 process separation in LAB would hold strong (which it did VERY well). Now I just need that stuff to come back over to RGB without a single trace of a surgery scar.

Smart Objects?

SRB Photo wrote:
Smart Objects?

Nope...doesn't work. tried turning them all to SmrtObjcts, and then converting to RGB and didn't work. Same result. This is discouraging because it seems the LAB method is the only way you can have 3 layers of info and not have it change anything...but LAB mode (for me) is a pain to work in for my retouching process as nothing is the same.

Elite Retouch wrote:

Nope...doesn't work. tried turning them all to SmrtObjcts, and then converting to RGB and didn't work. Same result. This is discouraging because it seems the LAB method is the only way you can have 3 layers of info and not have it change anything...but LAB mode (for me) is a pain to work in for my retouching process as nothing is the same.

I was suggesting to move the whole set of LAB layers into a single Smart Object; allowing you to keep them in LAB while the document itself is in RGB.

SRB Photo wrote:
I was suggesting to move the whole set of LAB layers into a single Smart Object; allowing you to keep them in LAB while the document itself is in RGB.

Oh lord no lol. I need all three to come back into RGB so I can retouch them. For some reason...healing and D&B on those layers isn't the same. I started healing on the detail layer...and those parts became super saturated. And then I tried doing my color adjustments, but the whole wheel is off (from what I'm used to in RGB) and I just could not get it right. Then I tried to D&B on my shaded layer...lord...I'm having nightmares and I'm still awake.

Elite Retouch wrote:
I started healing on the detail layer...and those parts became super saturated.

Now that's odd. What sample method did you select for the clone stamp and healing brush? It should be just "Current layer", neither "Current and below" nor "All Layers" (which could cause just that kind of problems).

Elite Retouch wrote:
I tried doing my color adjustments, but the whole wheel is off

Color indeed is a little different in LAB to what you're familiar with in RGB and CMYK.
Your starting with complementary colors (green-magenta in the a-channel and blue-yellow in the b-channel) with the neutral point of both channels being 0 and the cool colors on the dark side.
Furthermore you have to notice, that the color is completely separated from the luminosity, so trying to correct the shadows independently from the midtones won't work that easy in LAB.
Instead you have to correct for the shadows in one Curves adjustment and use the "Blend if" sliders to mask the midtones and highlights.
However the Color Balance adj. layer works great, so maybe that'll do the job for you.

Oh and one other thing could possibly be wondering you. Separating the color from the luminosity allows for imagenery colors (for example a vivid red that is totally black), so maybe changing the color will also result in toned highlights/shadows.

Peano wrote:

No need to change to LAB for that. Just change blend mode to luminosity in RGB.

This post in another thread just helped me to finally complete the image equation and achieve what I've been wanting. I'll post the completed action when I'm done.

Elite Retouch wrote:
Ok SRB, DerW...I got it working.

I wasn't too far off with having the bottom layer as a desaturated version of the blurred layer...but that's also what was causing the value shift. After visiting that one thread about LAB color space and exactly how it worked I wondered to myself if I could do to the luminosity (my supposed value only layer) what I did to get JUST the color info on a layer on it's on.

So I did it and BAM. No more color shifts for the 3 layer separation. So now my layers are in order as follows:

Detail (The original "Apply Image" layer)
Color (GB layer set to "Color" bm, and merged to a 50% gray layer set to "Color" bm)
Shading (GB layer set to "Luminosity" bm, and merged to a 50% gray layer set to "Color" bm)

Genius! Too bad I'm at work and can't play with it right now.

Elite Retouch wrote:
Detail (The original "Apply Image" layer)
Color (GB layer set to "Color" bm, and merged to a 50% gray layer set to "Color" bm)
Shading (GB layer set to "Luminosity" bm, and merged to a 50% gray layer set to "Luminosity" bm)

GB layer set to color doesnt seem like a good idea since blurring eliminates color detail. Try a higher blur and examine carefully.
I believe its better to use the original on gray...

So why dont we have a copy of the original layer on top set to color and use a desaturated (without the command) blur layer on bottom and HF in between?

Elite Retouch wrote:

This post in another thread just helped me to finally complete the image equation and achieve what I've been wanting. I'll post the completed action when I'm done.

When can we see that magic action:) sounds sweet as hell.

Koray wrote:
GB layer set to color doesnt seem like a good idea since blurring eliminates color detail. Try a higher blur and examine carefully.
I believe its better to use the original on gray...

Already tried. Not better.

Koray wrote:
So why dont we have a copy of the original layer on top set to color and use a desaturated (without the command) blur layer on bottom and HF in between?

Because I was getting a value shift this way. The way I did it is the only way it works as of now.

Elite Retouch wrote:
Here is the action!

WARNING: This is for 16bit editors only. It will not work for 8bit editors. Also I DO NOT import my image in as a SmrtObjct, and the import naming is the default locked layer "Background".

So if you get an error make sure you're in 16bit mode, not working on a SmrtObjct, and it's the original flattened "Background" layer.

http://www.sendspace.com/file/1u51uf

Thanks!!!! I will give it a try....

Elite Retouch wrote:
Here is the action!

WARNING: This is for 16bit editors only. It will not work for 8bit editors. Also I DO NOT import my image in as a SmrtObjct, and the import naming is the default locked layer "Background".

So if you get an error make sure you're in 16bit mode, not working on a SmrtObjct, and it's the original flattened "Background" layer.

http://www.sendspace.com/file/1u51uf

I just tried it and then I got to the "Fill" I would get a can't complete because of a program error. When I unchecked it from the action it completed and I ended up with a Separation DCL group with Detail, Color and Luminosity contained within it.

What was the Fill part of the action supposed to do?

Yes i was in 16 bit and RGB

ACPhotography wrote:

I just tried it and then I got to the "Fill" I would get a can't complete because of a program error. When I unchecked it from the action it completed and I ended up with a Separation DCL group with Detail, Color and Luminosity contained within it.

What was the Fill part of the action supposed to do?

Yes i was in 16 bit and RGB

As written, the action is trying to operate on a very large image (something like 16x20?) - resizing (w/o resampling) will allow you to do it, or you can just re-record the action using the Shift-F5 version of Fill vs. the paint bucket.

SRB Photo wrote:

As written, the action is trying to operate on a very large image (something like 16x20?) - resizing (w/o resampling) will allow you to do it, or you can just re-record the action using the Shift-F5 version of Fill vs. the paint bucket.

I unchecked the Fill that was in the action and put my own in there, working fine now...

I see the very slight color shift she was talking about, nothing unmanagable since I color correct last.

Definitely interesting and I think I'm going to spend most of the day playing with this one... I want to try it on a girl I have some pictures of with very bad skin, being able to attack the luminosity, detail and color separately may be a saving grace for retouching skin like hers... (Only time I've resorted to blur...)

ACPhotography wrote:

I unchecked the Fill that was in the action and put my own in there, working fine now...

I see the very slight color shift she was talking about, nothing unmanagable since I color correct last.

Definitely interesting and I think I'm going to spend most of the day playing with this one... I want to try it on a girl I have some pictures of with very bad skin, being able to attack the luminosity, detail and color separately may be a saving grace for retouching skin like hers... (Only time I've resorted to blur...)

There SHOULDN'T be a value shift anymore as the Luminosity padding corrected that. I have updated the action and will re-post it very soon.

DerW wrote:
I had some problems running it (although I had already recorded my own action, just wanted to see how you did it ;-)) on my German version of Photoshop.
This however could be avoided, if you used the shortcuts to select the bottommost, next, previous etc. layer (I guess that'd be [Shift]+[Alt]+[]] for the bottommost, [Alt]+[[] for the previous and [Alt]+[]] for the next layer... it's different in German versions, so I'm not sure about it).

However after I changed the language of PS to English it ran without any problems :-).

I didn't know this, however I have redone the action to accommodate this change and will be re-posting soon.

Elite Retouch wrote:

There SHOULDN'T be a value shift anymore as the Luminosity padding corrected that. I have updated the action and will re-post it very soon.

I've been playing with this and looking forward to seeing your improvements. I have also noticed some halos around edges.

But I find the detail, color and luminosity separate to be useful for more than just portraits...

Any particular reason why you default to 20 px for the GB?

ACPhotography wrote:
Any particular reason why you default to 20 px for the GB?

Because that seemed to be THE number when I was working on a lot of projects. I'd never published an action before so there are many things I had no idea to do lol.

If the number is too big for particular project, you can reduce it I believe.

Elite Retouch wrote:

Because that seemed to be THE number when I was working on a lot of projects. I'd never published an action before so there are many things I had no idea to do lol.

If the number is too big for particular project, you can reduce it I believe.

I actually kept running the action and deleting it on the same picture at different radii just to see the difference. I did find that with the 12mp out of my Nikon, 20 was too high, I'm having some negatives scanned which will be much more than 12mp so I'll see how the 20 does on those. I have some 17mp from a leaf/mamiya MF I should try it on.

Elite Retouch wrote:
For those of you that were having problems, here's the new action made using ALL HotKey commands. It was very tough as one mistake and I was back to square one.

Here you are: http://www.sendspace.com/file/6h880q

Works perfectly, I know all about making actions I can't tell you how many times I've had to restart from the beginning...

ACPhotography wrote:
Works perfectly, I know all about making actions I can't tell you how many times I've had to restart from the beginning...

Yeah. I think the next major thing Adobe should implement into PS is the ability to have TOTAL CONTROL of that actions pallet. It's so constricting when it comes to changing SLIGHT mistakes.