What is "Apply Image" really doing?

Pindy

I'm going through Kelby's 7-Point book, and thought I would try to use the switch to LAB mode then Apply Image procedure on an image of my own. It worked brilliantly for this image, but I don't have any idea what it's doing when you "apply image". You've converted the color mode to LAB, so what image does it actually apply?

jjbong

Pindy wrote:

I'm going through Kelby's 7-Point book, and thought I would try to use the switch to LAB mode then Apply Image procedure on an image of my own. It worked brilliantly for this image, but I don't have any idea what it's doing when you "apply image". You've converted the color mode to LAB, so what image does it actually apply?

I am not familiar with the book, so I can't give you a compact answer. If you post the steps involved, I probably could.

Briefly, Apply Image modifies the target channels (the ones currently selected) by using a transform of the source channel (both the transform and the source channel are selected in the Apply Image dialog), and possibly a mask.

Rather than attempt a long discourse on this, I'll give two examples.

The first is that I want to replace the Red channel in an RGB image with the Green channel (why I might want to do this is another issue). I would select the Red channel, and then Apply Image selecting the Green channel in the current layer and Normal Mode (Normal means replace everything, just as it does in Layers).

The second is that I want to replace the Red channel in an RGB image with the Green channel, but only in areas where the Green channel is darker than the Red. I would select the Red channel, and then Apply Image selecting the Green channel in the current layer and Darken Mode.

I hope this helps, but I fear it might not. If you post the steps in the Kelby procedure, I'll do better.

BinaryFx

I agree with John, more would need to be known about the source and target channels and blend mode, opacity etc.

I have some links to the general topic here:

http://members.ozemail.com.au/~binaryfx/links.html#C

Scroll down to the "Channel Editing, Channel Mixer, Calculations & Apply Image" links.


Regards,

Stephen Marsh
http://members.ozemail.com.au/~binaryfx

Pindy

Thanks all.

Kelby simply outlines the procedure as follows:

1. Switch to Lab mode.
2. Apply Image. Use Soft Light blend mode.
3. choose between Lab, a or b channels, whichever gives your image a nice color enhancement.
4. Vary the opacity if needed.
5. OK and back to RGB.

That's it.

jjbong

Pindy wrote:

Thanks all.

Kelby simply outlines the procedure as follows:

1. Switch to Lab mode.
2. Apply Image. Use Soft Light blend mode.
3. choose between Lab, a or b channels, whichever gives your image a nice color enhancement.
4. Vary the opacity if needed.
5. OK and back to RGB.

That's it.

A good reference is Dan Margulis' "Photoshop LAB Color" Chapter 15.

Here's what's going on.

You have all channels selected (L, a, and b). When you do the Apply Image, it blends the source channel (Lab, a, or b) into each of these channels in Soft Light Mode. Soft Light is a variant of Overlay. All Overlay variants will change channel based on the source channel as follows. If the source channel is 50% grey, there is no change. If the source channel is lighter than 50% grey, it lightens the channel. If the source channel is darker than 50% grey, it darkens the channel. How exactly it lightens or darkens depends on the mode. Soft Light is "gentler" than Overlay.

If you select Lab, it blends the L with the L, the a with the a, and the b with the b. The L to L blend increases contrast in the mid-tones. The a to a and b to b blends increase color contrast, or saturation. This particular combination results in a pleasing effect for many pictures.

If you select a or b, it is more complicated. Let's use a as an example. Recall that in Lab, a is 50% grey for neutral colors, lighter than 50% grey for colors that are more magenta than green, and darker than 50% grey for colors that are more green than magenta. This means that the blend will do the following:

nothing where the colors are neutral
darken where the colors are green
lighten where the colors are magenta

For the L channel, this is literally the effect you see - lightening greens and darkening magentas. For the a channel, this lightening and darkening increases saturation in the greens and magentas. For the b channel, it's more complicated. Anything green gets more blue, and anything magenta gets more yellow. The net is that the colors get a bit more intense, but also shift a bit.

This is a pretty terse summary, as there's a lot going on. If you understand the Lab color space, it shouldn't be too hard to follow.

arodney

Pindy wrote:

Kelby simply outlines the procedure as follows:

1. Switch to Lab mode.
2. Apply Image. Use Soft Light blend mode.
3. choose between Lab, a or b channels, whichever gives your image a nice color enhancement.
4. Vary the opacity if needed.
5. OK and back to RGB.

That's it.

My first question would be, WHY?

Pindy

arodney wrote:

My first question would be, WHY?

Thanks for playing. Cause it looks groovy, of course.

arodney

Pindy wrote:

Thanks for playing. Cause it looks groovy, of course.

Oh, you don't have the groovy filter installed?

Seriously, what's the point here?

jfriend

arodney wrote:

Oh, you don't have the groovy filter installed?

Seriously, what's the point here?

Andrew, I don't understand why you attempt to shoot down every single technique that involves using LAB. If you work in 16-bits, no harm is done to an image with a round trip to LAB and there are many things that are easy in LAB, particularly when you're trying to separate color from contrast or isolate specific colors or create color-specific masks or blend-ifs. I'm not arguing that they can't be done some other way. LAB is just a tool, not a religion someone should be for or against. People should use the tool that works best for them.

This particular technique does a nice subtle color enhancement. It's structurally similar to a curve applied to the A, B and L channels. There probably are twenty other ways of doing something that gives a similar, though not identical, result in Photoshop.

Pindy

More to the point, it does it in a minimum of steps.

arodney

jfriend wrote:

Andrew, I don't understand why you attempt to shoot down every single technique that involves using LAB.

How can I shoot it down until I know what the process is supposed to do?

I tried the steps on my Printer Test File. That image has preferred color appearance. I didn't see anything useful the process did to that kind of image. So I have to assume whatever you are describing, (realizing I don't have the book you reference), some kind of image correction. What's this supposed to do? Is this an unreasonable question to ask?

jfriend

arodney wrote:

How can I shoot it down until I know what the process is supposed to do?

I tried the steps on my Printer Test File. That image has preferred color appearance. I didn't see anything useful the process did to that kind of image. So I have to assume whatever you are describing, (realizing I don't have the book you reference), some kind of image correction. What's this supposed to do? Is this an unreasonable question to ask?

Some examples:
http://www.planetnikon.com/forums/index.php?showtopic=7303&mode=threaded&pid=40901
http://www.dgrin.com/showthread.php?t=24919
http://p213ae.files.wordpress.com/2008/01/arch-213-photoshop-post-production-techniques.pdf
http://blogs.tech-recipes.com/qmchenry/2008/02/01/lab-color-enhancement-central-park-wysteria/

arodney

jfriend wrote:

Some examples:
http://www.planetnikon.com/forums/index.php?

Is that the only image? The before isn't so great (the sky look pretty darn magenta) but the after is pretty ugly at least to me. Yes the sky is better hue, the rest of the image is pretty ugly. In a nutshell, when would I use this?

http://blogs.tech-recipes.com/qmchenry/2008/02/01/lab-color-enhancement-central-park-wysteria/

There's talk of sharpening, then a saturation tweak (which looks pretty awful to me). Nothing special I see here and something that was discussed as being just a rendering tweak one could apply in a good Raw processor like Lightroom.

Pindy

I think I used it in an altogether more benevolent way, though there are other processes at work here:



You may not dig it, but this ain't the Whipping Post, so I'm not really bothered. The fact was, that I couldn't get the kind of contrast and color out of Lightroom that I wanted in this particular photograph, without saturation increases beyond what I wanted. I'm kind of sponging at the moment, so whatever works, I'm open to it. Never should have bought that Wacom.

This is the default RAW conversion in LR:

Duffy Pratt

Andrew:

The technique, using overlay instead of soft light mode, is in one of the last chapters of The Canyon Conundrum. It's a recipe that Dan used, primarily for faces. When you apply the A or B channel to itself in overlay mode, you are driving the colors apart, and doing it in a way that becomes even more intense for the stronger colors. Neutrals, of course, stay neutral.

I've found this technique to be useful when you have a portrait with flat looking skin tones. This is one area where I haven't been satisfied with any of the RGB saturation tools. Vibrance doesn't hit skin tones, so its more or less useless here. Saturation in ACR is too hamfisted, in my hands at least, to do a good job on faces.

I know there are people who do great work with skin tones without resorting to LAB. But when you only have 3-5 minutes to work on a picture, and you want to get some bang for you buck, I think this is a pretty neat trick.

Hope this helps.

Duffy

jjbong

Pindy wrote:

I think I used it in an altogether more benevolent way, though there are other processes at work here:

Nice job. This is just the kind of shot where a bit of LAB can help. Fairly dull tones overall, which gives a lot of room for LAB to do what it does best without introducing color artifact. I assume you selected the Lab option, as the overall contrast is boosted quite a bit (and favorably).

My only critique is that the dock is a bit over-saturated. If you agreed with that judgement, you could easily handle it in your workflow with some blend-ifs in LAB after you did the Apply Image.

My workflow is quite different, and the Kelby 7-step wouldn't fit into it. But I certainly wouldn't cast asparagus on it, and you got a nice result out of it.

I completely agree with jfriend's post earlier on the thread. There are "tools" available with LAB which are very useful for certain types of problems, just as there are lots of other tools available for other types of problems. It's not a religious issue (although some of the posts on this thread have an emotional intensity that suggest that it might be for some). It's just a matter of getting the result you want as efficiently as you can. The key to all of this, in my view, is understanding what you're trying to get out of the image (what you need to fix, what you saw shooting that the camera didn't, etc.), and selecting the best tools for the job. The time constraint is a factor, also, as Duffy points out.

BinaryFx

I am a big fan of channel blends, even if they do not promote raw processors...perhaps in the future if a raw converter can do this it will then be touted as a reason to purchase/upgrade. Who knows, perhaps some existing raw conversion software currently does apply image type moves under the hood without users being aware of it.

Back to apply image. As we have seen, it is possible to mix colour modes, making use of both RGB in one file and Lab in another file, blending the two files together using various source and destination channels and blend modes, opacity etc. This is very powerful stuff, I was probably first introduced to this general topic many years ago when Kai Krausse was using ChOps for special effects in his Power Tips, back before layers existed in Photoshop.

The default rendering from the raw image that Pindy posted makes a good example for another apply image move, using the third commonly available colour mode in Photoshop - CMYK. As this is a raw file, one can obviously render out a more appealing version than the default. This post is more about adding to the original topic of apply image, whether one can do similar with other tools is a side issue perhaps best explored in a different topic thread.

Steps and explanatory notes. It always sounds harder/longer than doing it or recording an action so don't be discouraged before trying it:

1. Duplicate the current image and convert it to CMYK (flattening).

What CMYK you may ask? That will depend on what artisic rendering you are going after! For this example, I will simply use the US Coated SWOP v2 profile. That being said, for this following method - custom CMYK is a better choice, as one can create a tailored separation with the desired characteristics rather than the fixed conditions found in a true ICC profile. The work around is to try many different CMYK profiles, if not using the more flexible Custom CMYK interface. The intent will be to use the conversion in a channel blend via the apply image command.

2. Keep the CMYK image open and return to the main RGB file. Dupe the flat background image into a new layer.

3. Call up the apply image command. The source document is the CMYK docs flat background layer. The source channel is the Black channel. The blending mode is Multiply with opacity at 100%.

4. After applying the K channel to the duped RGB layer, one should have darker shadows and more density in the three quarter tones to the midtones. One can adjust global opacity or use a grayscale layer mask or blend if sliders to vary the tonal blending of the darker data into the lighter original data. Another option is to curve the K channel before blending. There are many possibilities!

A very quick example is found below in the right hand section of the image. This is a 100% opacity multiply blend of the K into the RGB. I used blend if sliders to restrict/reduce the effect in the deep shadows, while having full effect in the midtones to highlights.

There are other ways to achieve similar visual results, with or without using the K channel from CMYK (that is the great thing about Photoshop, so many ways to do things). It will depend on the image and the operator as to which particular method and result may be preferred. The important general concept to take away from all this is that the different colour modes all have different channel characteristics that can be used with blending modes via the apply image command to alter an image in various ways.


Sincerely,

Stephen Marsh
http://members.ozemail.com.au/~binaryfx/

arodney

Pindy wrote:

This is the default RAW conversion in LR:

Not useful. How about showing the best desired rendering you can muster? Or upload the Raw.

Showing the default rendering and then the "after" using the Lab technique is like showing the default rendering and the same image in Photoshop prior, what's the point?

jjbong

BinaryFx wrote:

A very quick example is found below in the right hand section of the image. This is a 100% opacity multiply blend of the K into the RGB. I used blend if sliders to restrict/reduce the effect in the deep shadows, while having full effect in the midtones to highlights.

Stephen,

I'm curious how you did this. I've not had much success with blend-if in RGB. I usually convert to Lab for that.

jfriend

arodney wrote:

Not useful. How about showing the best desired rendering you can muster? Or upload the Raw.

Showing the default rendering and then the "after" using the Lab technique is like showing the default rendering and the same image in Photoshop prior, what's the point?

Andrew, it is not someone else's responsibility to prove to you that a result done with a LAB technique cannot be achieved in a RAW processor. All he has to do is prove to himself that he can get a result he likes with any technique he knows. In this case, he likes the result he achieved using LAB techniques and he knows how to use those tools and he decided to share. What is wrong with that? There are probably hundreds of different approaches that could be used on this photo. The fact that 99 other approaches exist doesn't making any one more or less "correct". Some, I'm sure, would do more in the RAW processor first. Some would do less.

You act like anyone who does something outside a RAW processor (particularly in LAB) owes you a proof that it can't be done in the RAW processor in order to justify any other technique. That's not his responsibility. If you want to share that more could be done in the RAW processor or that better results could be achieved with different techniques, then ask for a copy of the RAW image and share your technique with it rather than attempt to poke holes with what's being shared here.

Your tone in this thread is very clear to all, so don't go saying you're just asking questions and you don't understand.

BinaryFx

jjbong wrote:

Stephen,

I'm curious how you did this. I've not had much success with blend-if in RGB. I usually convert to Lab for that.

John, the blend if move is using the "gray" or RGB luminosity data of the image (not the same as the L of Lab, although visually similar). As this is a tonal based move, adding darkness to all of the tonal range except the deep shadows - blend if gray in RGB has no problem and is similar to doing things with blend if to the L in Lab mode. Additionally, the nature of a good K plate provides a nice near seamless blend back into the RGB. It is my guess that the advantage that your refer to in Lab blend if method is likely to do with the A and B channels more so than the L.

From memory, the blend if was rather simple, probably something like splitting the shadow slider from 0 to 15/65 for both the "this layer" and "underlying layer" layer option blend if gray (luminosity) sliders.

More on blend if here:

http://members.ozemail.com.au/~binaryfx/blendif.html
http://members.ozemail.com.au/~binaryfx/links.html#B



Regards,

Stephen Marsh
http://members.ozemail.com.au/~binaryfx/

Pindy

Stephen—will look into that technique. Always good to learn something new. Amazing how many avenues there are in Photoshop.

Oddly enough, most of the contrast in the image came from the LAB Apply Image. I have JP Caponigro to thank for the light-painting.

As for Andrew, below is a link to the RAW file. I'd love to see what you can do just in a RAW processor or indeed in any editor. If you'd like to really be helpful, sort out this image in Lightroom and post an XMP so I can see what the deuce you've done to it, GRAB IT HERE

Greenwich.CR2

Pindy

arodney wrote:

Not useful. How about showing the best desired rendering you can muster?

okay then. It's LR only:



I still like my original better. Trust me, I'd love to do it all in Lightroom, and most of the time, I do.

arodney

jfriend wrote:

You act like anyone who does something outside a RAW processor (particularly in LAB) owes you a proof that it can't be done in the RAW processor in order to justify any other technique.

Only down on the speed and image degradation and the idea of fixing something that isn't broken to begin with. Other than that, I think its great.

arodney

Pindy wrote:

As for Andrew, below is a link to the RAW file. I'd love to see what you can do just in a RAW processor or indeed in any editor.

Downloading now, will play and upload something.

arodney

arodney wrote:

Downloading now, will play and upload something.

Here's a 100% LR rendering and the Lab technique. Not identical pretty close and took about 3 minutes (the biggest job was trying to keep a rendering that matched what you did, rather than what I'd do). Certainly a far cry from the LR defaults and from your attempt at using the various LR rendering controls. All in all, I think I'm pretty close. Had I not tried to match the original (yours). It would have been accomplished in seconds by pulling on a few sliders. Best of all, I have the entire edit list forever, I can go back and make tiny tweaks or begin from a Vertical Copy prior to ever building a single rendered pixel.

http://digitaldog.net/files/Greenwichtest.jpg

arodney

An inline of just the LR version:

arodney

jfriend wrote:

Your tone in this thread is very clear to all, so don't go saying you're just asking questions and you don't understand.

I don't have Scott's book (does everyone else reading this thread automatically assume to have it?). The technique's goal was never discussed nor defined until I asked (after using the technique on an image that is in a desired color appearance). If you want to go on record based on the above facts that I did or didn't understand the clarity of the technique, I might suggest you are way off base bud.

Also, thanks for stating your opinions of my tone for all readers ("its very clear to all").

Pindy

Nice. Again, an XMP would make this a "learning experience".

If you're so inclined, feel free to do what YOU would do. I found this particular image to be challenging and my interpretation of what I saw that morning is only that.

Thanks.

arodney

Pindy wrote:

Nice. Again, an XMP would make this a "learning experience".

I'll shoot you an email where I'll upload a DNG. Since its your image, unless you say otherwise, I don't want to post it publicly, but you'll see all the settings in ACR or LR.

arodney

Pindy wrote:

Nice. Again, an XMP would make this a "learning experience".

If you're so inclined, feel free to do what YOU would do. I found this particular image to be challenging and my interpretation of what I saw that morning is only that.

Thanks.

I can't. I wasn't there, didn't shoot the image. The goal here was to make a version that was pretty close to yours without going nuts. Note my rendering is a bit warmer. I see the biggest differences in the water in the foreground.

The point was, its a heck of a lot harder to attempt to mimic something than just render it as you would as you work on your own stuff. Check out the Podcast by Jardine on subjective rendering in Lightroom. He does the work quickly because he knows the tools and what he wants.

http://lightroom-news.com/2008/02/25/lightroom-tutorial-podcast-51-has-been-posted-subjective-color-correction/