I couldn't help practicing Dan Margulis' Picture Postcard Workflow on divamum's harsh light portrait. So, with her permission, I'm going to share this process and results here.



I started out just to show how I'd dodge the face, but quickly realized that I could get a much better result by including this move within the whole workflow. So here goes. If you get bored with the steps, you can skip ahead and read my conclusions at the end.

Step by step

1. Adjust casts with RGB curves in color blend mode When I opened this image, I measured the colors with the color sampler and discovered a slight cold cast. This is especially true of the wall in the background which measures a bit blue-green (A and B both positve). This is probably due to the mixed cast with the sunlight and shadow. The subject is almost all in this shadow, so fixing the wall would probably also be good for the face. I used these curves to make the wall mostly neutral, warming up the midtones of the whole image.
   
   
   
   One of the keys to the PPWF is to use color blend for this curve, which means you don't have to worry at all about its effect on contrast. You just get the color adjustment. This makes it tons and tons easier to write the curves. The result is subtle, but a definite improvement to my eye.
   
   
   
   The reason to do this first thing is that as future steps enhance colors, we'll be starting with colors that are our friends and so can stand to be enhanced. If we left that slight cold cast, we'd end up with a very cold image once we enhanced its colors.
   
   
2. Skip channel blending Normally in portraits, we'd just use the green channel for luminosity as the first step in contrast enhancement. But this image has more than enough contrast on the subject's face, and we don't want more. Perhaps we could use an unconventional move like blending the red into the green to flatten her face somewhat? But this is supposed to be a fast workflow, so forget it. We plan to dodge a bit later to fix this problem, so no action required here. On to the next step.
   
   
3. Adjust contrast with RGB curves in luminosity blend mode The idea here is much simpler than trying to use the same set of curves for both color and contrast. Now we can just examine both the red and green channels (blue doesn't count for enough contrast to matter). We'll just adjust the endpoints of the curves to delineate the darkest and lightest significant features of the image and then pull the shape the curves to make them steep through the parts of the image we care most about. Put the curves layer into luminosity blend mode before evaluating the effect. I am finding that this technique gives really nice contrast enhancement without tending to make the image too harsh.
   
   
   
   
   
   
4. Dodging with a screen blend layer, layer mask, and big, soft, low opacity brush This step is definitely not a conventional Dan Margulis. He generally will do a lot of work to avoid local adjustments like burning/dodging. Perhaps a subtle use of shadow/highlight or the like would help, but our problem is really in the 3/4 tones. So when you have to dodge, well, you have to dodge. There might be a clever alternative, but mindless is quick and effective.
   
   You could use the dodge tool, but it's awkward because the tools seem to have a mind of their own, aimed at only highlight, midtones, or shadows. This technique is a very direct alternative. Make a dupliate layer. Set blending mode to screen. The result will be very light. Add a layer mask and use Edit->Fill to fill it with black.
   
   
   
   Use a large softest brush to paint white on the layer mask with very low opacity (say 10%). I clicked 5 times on her face to get this.
   
   
   
   
5. Save a duplicate of the contrast enhanced version We'll want to consider blending with it later to soften the results of the following steps in case we get too much color or sharpening harshness.
   
   
6. Color enhancement in with channel blending in LAB The next two steps will seem very odd at first and understanding how they work in detail requires understanding all four of LAB, overlay blends, layer masks, and the Apply Image command. I suppose that makes them advanced. On the other hand, they are quick to do and could easily be distilled into actions to make them even quicker. I recommend trying this out without thinking about it too much at first. Then you can always learn more about the relevant tools and ideas.
   
   Convert the image to LAB, flattening if still necessary. Make a duplicate layer in Normal mode. Use Image->Applied Image to apply the image to itself in Overlay mode.
   
   
   
   The result is very colorful and very harsh. Don't worry about that yet. We didn't really want to overlay the L channel with itself, at least not this much. So we'll use Apply Image again to restore the L channel from the background. We could use 100% to restore it 100%, but adding a little more contrast to match the enhanced color is often a good idea. Play around with values between 80 and 100%. I chose 85%
   
   
   
   At this point, the image is starting to look pretty good, if a bit over the top:
   
   
   
   
7. Use a channel as a layer mask to control the blend The big idea here is that we want more color enhancement were the image is lighter and less where the image is darker. This is because we don't see color as well in darkness as in light. Dan has a lot to say about this in various places. For our purposes, we'll just see that it works and perhaps discuss later.
   
   Add a layer mask to the that overly bright top layer. Use our friend Apply Image to apply the L channel to the layer mask (select the layer mask and then invoke Apply Image):
   
   
   
   Dan suggests applying a small blur to this layer mask, say radius 4.
   
   Anyway, big improvement:
   
   
   
   
8. HIRALOAM USM sharpening in all three LAB channels. This isn't a PPWF step per se, but we'd like this to look sharper so I went for it. It seems that some conventional sharpening has already been done, so I skipped that part and used USM 28%, 15.9, 0 to add some local contrast and color variation. Search dgrin for HIRALOAM and you'll find lots of discussion of this.
   
   
   
   
9. Adjust to taste The only thing left to is to decide just how much we want to keep of that color we just introduced. That's why we saved away a duplicate before we went to LAB. Now convert the LAB verson to RGB, layer that duplicate on top, and adjust the opacity. I liked about 50%. Fini!
   
   

Conclusions

Hmm, 9 steps, kind of a lot to absorb. Nice result. Might be much easier in LR? (If so I'd like the same sort of careful description I just gave.) The steps in this workflow come pretty quickly once you learn them, but learning will take some practice and study.

I really like separating RGB curving into two steps, one for color and one for contrast. Seems to give good results with little work. The contrast enhancement gets a lot of contrast without harshness. I've known how to use color blended curves to correct casts for a long time, but this workflow makes it systematic.

So, this wasn't a picture postcard, and I didn't go for a picture postcard result. But the Picture Postcard Workflow worked we..

Ummm... WOW. This actually looks better than the reality did....

I should have sent you the raw file to work from - sorry I didn't think of that sooner (it was shot raw, although the SOOC posted that I think you worked from was converted to a jpg for posting).

I have to read this about a million times to try and understand what you did - there are some methods you've used within each step which are still way over my head, but the results demonstrate it's clearly worth coming to terms with them!

Thanks for sharing this - I can't wait to dig in and learn how to DO this for myself!!

ETA: I responded in the original thread with the picture, but if you want a raw file just let me know! (It's ~14mb, so let me know how to get it to you if you want it). That said, the converted jpg doesn't seem to have slowed you down or caused you any problems, at least at that resolution!

I received divamum's RAW file and I took the liberty of forwarding it to rutt as well.

I opened the RAW file in Adobe RAW converter for CS4 - I believe it is the current version 5.3 perhaps?



I begin by choosing a camera profile that looks "better" on my calibrated screen - click on the third box from the right for Camera Profiles - I chose Camera Standard - I find this choice works rather well for people and snapshots. For landscapes I tend to prefer Landscape - well duh!!


After an appropriate profile, I began by looking at the color with the white balance eye dropper and reading the pixels in the wall behind her head, in the black fence, her shoes, the sidewalk, to begin to get an idea where to find a suitable point to balance with. Usually white is an excellent choice, but although we "see" that her dress is white, it is really a very light, bright blue, because if you try to balance on the dress, the image does not work.

Next is color balancing - (the exposure settings need to be close to correct before finalizing color balance settings.) I used the eye dropper on several areas in this image - there are numerous possibilities - the "grey?" wall behind her, the black fence, the black lamp post, the "grey?" sidewalk behind her - in the shade or in the sunlight?

I chose to click on back of shoe for white balance, rather than the wall behind her head - I am not sure that wall is really a neutral gray, but her black shoes probably are. The Black fencing might also work but the shoes seemed to look better overall for me. My color settings finally were Custom 5650 for Temp and - 3 for tint.

This gives pretty decent skin tones. Here anterior axilla reads 167, 136, 103 which isn't that bad for Caucasian skin. The wall reads ~ 158,158, 155, close to grey but not really.

Exposure +0.05 - exposure was spot on!

Recovery +79 That's more than usual but a lot was needed to regain the blown areas along the viewer's right side of her dress in the sunlight

Blacks - +22 That is higher than I usually use also, but necessary to get a black in the bow in her hair and the fence was still not pegged

Fill lIght +23 to lighten her face

I also increased the Brightness to +65 to lighten the mid tones as well.

I reduced the contrast slightly at +22.

I added a bit of Clarity with + 10 -- Jeffe Schewe says almost every image can benefit from a touch of Clarity and who am I to argue with him.

Vibrance I set to +18 and Saturation I left at Zero (0)

The Tone Curve I left at the standard of Medium contrast. I could add a bit of contrast and lighten her face by steepening the curve in that area, but chose not to at this time.

I tend to sharpen a bit more than some folks. I set the Amount at 78, the Radius at 1.0 ( I rarely change it ) the Detail at 32, and the Masking I set at 35 ( I frequently have Masking anywhere between 30 and 70% which really limits the sharpening to major edges in the image)

Luminance was left at 0 since this image was shot at ISO 100 and did not seem to be noisy. Color Noise I left at 25 which is the default.

I did not punch any colors in the Hue/Saturation/Luminance columns, nor did I do any Split Toning.

Chromatic Correction is a function of what specific lens was used. For this image I set Red/Cyan at +10, Blue/Yellow at -15, and Defringe to All Edges. I use Defringe All Edges routinely, and correct Red/Cyan and Blue/Green as needed as I inspect the image at 300% near the corners of the image.


At this point, I am basically done with my setting for Adobe Raw Converter, unless I see something I need to go back and alter.

I chose to open this image in ProPhoto RGB as a 16 bit image 12.2 Mpxles and 300 pixel per inch or 2848 by 4272 pixels


I opened this image in Photoshop, sharpened slightly with Amount 22, Radius 50, and Threshold 2 for a bit of local contrast enhancement and edited with a luminosity blend at 90%. I converted to 8 bits, sRGB, and saved the file as a jpg level 10 and uploaded it to my smugmug account. It can be seen here - http://pathfinder.smugmug.com/galler...60624351_ZtjuE and the original full size file is available for inspection.



John's image has a bit more punch in the pink colors in the dress - I could have done that in ACR but did not at this time. To run through my editing steps in ACR normally takes just a couple minutes - longer to record and describe than to do.

Bear in mind, this image I am posting has had no significant local image editing in Photoshop - this is basically straight from RAW.

WOW, thanks for posting the steps for that, I really needed to learn how to do that, I am almost embarrassed to admit. The photo looks awesome, I can't wait to start practicing. This will undoubtedly make a big difference in my finished product. I LOVE this place!

Thanks again to everyone that took the time to instruct some of us!!

This is so fascinating to see what you experts are doing... with my own image! The results are fantastic and so wonderfully helpful to me to see HOW you do what you do.

A couple of things I'll throw out here (having, yanno, been there in person ):

- the spots on the dress really are bright pink, and the bg really is white (whether "true" numerical white I can't say, but it's a white dress with bright pink dots). Rutt's final colours on that are closer to the reality of it, at least on my (Huey-calibrated) laptop monitor.

- the corrections that both of you have made actually result in the photograph having MORE pleasing light and tonesthan the reality. Although it was 5pm, the sun was still high enough that it felt bright and uncomfortable (one of the reasons she's positioned as she is - if I turned her through 180 degrees she complained of the bright sun in her eyes). And I'd say that blue-ish cast was real - I can only assume it's something to do with reflections from the surrounding area, where there is a small park (green grass) , a large monument/statue in black, and a lot of grey-er stone buildings than the one which is visible.

- what you both prove is that we can not only "salvage" an image in tough lighting, but with the right skillset and approach actually make it look GREAT... and still entirely natural.

Thanks again - I'm so grateful you both took the time to work on this image!!

Let's take a side-by-side comparison of the results here.

• Differences are subtle but real.
• We both warmed the image somewhat, with positive results. Warming moved from midday to sweet light.
• I was more conservative about adding color pop. You can see this best in the face, lips especially. This is a question of taste, but the fact that I dialed back all the color in one final step might have made this easier to control.
• I can really see the effect of using the L channel as a layer mask for the color enhancement move. Much less color enhancement in mine in the shadows which I think adds to the sense of realism in spite of the improvements. Look under her eyes, for example. Or at the fence.
• Dodging the face (not at all a part of Dan's workflow) seems to have been a little more successful in removing the harsh shadow/light from the shot than the LR steps. Crude but effective.
• I got more contrast in her face without making it as harsh. I think this came from shaping the R and G curves to make them steep there.
• Can barely see the differences in sharpening and/or HIRALOM

So, all in all, I have to score it a modest win for the PPWF in terms of quality of the final result. Of course, both Jim and I are more responsible for that then the tools we were using. Jim can probably easily scale back his color enhancement just as I could increase mine. Using the L channel as a mask for color enhancement is a real visible win, probably not that easy to get with a pure LR workflow at present.

LR probably took less thought/time than the PPWF. But one advantage of a well structured workflow is that it doesn't take a lot of time either once you learn it. There are set point for each decision and the relativly simple criteria for making them. Using a few actions, things go very fast. Dan's advertised goal for the PPWF is about 4 minutes per shot, comparable with the time Jim spent, though the overhead to learn is probably longer.

Firstly, my thanks again to both of you. I hope I didn't imply there was some kind of "competition" between the two versions - they both have much to offer, and it is beyond awesome to see how you're each approaching it with different methods!!

I will say that for me, at this stage of development, the LR way would be MUCH easier to duplicate. I absolutely LOVE your posts, but they often use sophisticated elements and functions of PS which I don't "get" yet, so I have to laboriously work through, figure out the parts that you haven't explicitly explained (and thanks for the screenshots - they SERIOUSLY help with finding one's way around some of these "buried" features in the software) and then try to figure out how to use it

One thing I note that both of you often reference is the numerical values of specific colours. However, I never know quite what those numbers are referring to - is it a theoretical "colour", or the same as web colour numbers, or is there a chart somewhere I should have at my fingertips to look and compare? I suspect this is a very basic colour management/Photoshop skill, and one that I have completely missed somewhere along the line...

Thanks again so much for taking the time to show us all how it's done.... using EITHER workflow!

Originally Posted by divamum One thing I note that both of you often reference is the numerical values of specific colours. However, I never know quite what those numbers are referring to - is it a theoretical "colour", or the same as web colour numbers, or is there a chart somewhere I should have at my fingertips to look and compare? I suspect this is a very basic colour management/Photoshop skill, and one that I have completely missed somewhere along the line...

The numbers are the actual values of sampled pixels. These can be seen in the Info palette when you hover the cursor over any area. You can set up the palette to display multiple color spaces as well as before/after values when you are doing corrections. You can also set up a number of fixed position samples.

Originally Posted by Richard The numbers are the actual values of sampled pixels. These can be seen in the Info palette when you hover the cursor over any area. You can set up the palette to display multiple color spaces as well as before/after values when you are doing corrections. You can also set up a number of fixed position samples.

Well, I know how to find out what they are by using the dropper and reading the info pane, but I'm still not sure what they're *used* for, if that makes sense. When I see people discussing the numbers I have no idea what they reference.... or why we need those numbers. Are there specific numbers we "should" have for any given colour and against which we "should" measure the colours in our own image? This is the part I'm having difficulty understanding.... (I suspect it's very simple, I just haven't had that "aha" moment yet - I feel a bit like I'm back in 9th grade algebra waiting for the penny to drop!)

Hitting the i ( eye ) key in Photoshop brings up the color picker sampler tool and this will read the pixel values, either of a single pixel, 3x3 average, or a 5x5 average ( CS4 offers even more choices ) which is selectable in your tool bar and the data are visible in the Info palette (In CS4 you must click on the Info palette to see it.) In CS3 it was easier to see in my opinion - one of the upgrades of CS4 that I am not too fond of.

If the Info palette is not visible, you may need to go to Window -> Info or hit the F8 key in CS4.

Once you do this, you can read the individual pixel data in RGB, or LAB, or CMYK. It helps to know that black in RGB will look like 5,5,5 or 7,7,7 and white will look like 248,248,248 - all three channels are equal or close to it. In LAB, black and white will have a and b values near zero, and an L value very near zero for black, while white will have a value near 100 or so. Greys will have a and b values near zero, and an L value near 50.

Once you get used to reading pixels like this, finding neutrals becomes easy in many images ( but not all )

Originally Posted by rutt Let's take a side-by-side comparison of the results here. Differences are subtle but real. We both warmed the image somewhat, with positive results. Warming moved from midday to sweet light. I was more conservative about adding color pop. You can see this best in the face, lips especially. This is a question of taste, but the fact that I dialed back all the color in one final step might have made this easier to control. I can really see the effect of using the L channel as a layer mask for the color enhancement move. Much less color enhancement in mine in the shadows which I think adds to the sense of realism in spite of the improvements. Look under her eyes, for example. Or at the fence. Dodging the face (not at all a part of Dan's workflow) seems to have been a little more successful in removing the harsh shadow/light from the shot than the LR steps. Crude but effective. I got more contrast in her face without making it as harsh. I think this came from shaping the R and G curves to make them steep there. Can barely see the differences in sharpening and/or HIRALOM So, all in all, I have to score it a modest win for the PPWF in terms of quality of the final result. Of course, both Jim and I are more responsible for that then the tools we were using. Jim can probably easily scale back his color enhancement just as I could increase mine. Using the L channel as a mask for color enhancement is a real visible win, probably not that easy to get with a pure LR workflow at present. LR probably took less thought/time than the PPWF. But one advantage of a well structured workflow is that it doesn't take a lot of time either once you learn it. There are set point for each decision and the relativly simple criteria for making them. Using a few actions, things go very fast. Dan's advertised goal for the PPWF is about 4 minutes per shot, comparable with the time Jim spent, though the overhead to learn is probably longer.

As I look at our images, I agree that there are differences, some of yours are better, but I am not sure all are.

Your dress is bluer than mine, and you have more blown highlights along her left side in the sunlit area. You have more posterization in the light glow behind her. I am not saying these are better or worse, that is a value judgement based on your desires for the image. I specifically tried to avoid the frying the highlights on her dress, and could have held more contrast in her dress if I had not I think. My lamp glow is not as posterized. Not better, but different.

I did not use any masking for my image, nor did I tackle a try with an adjustment brush in ACR either ( it was after midnight last nite when I ran through that edit ) but that could still be done. I did not punch the saturation of the pink in the dress, but the magenta slider in ACR certainly would have increased that, as would a different camera profile choice.

If I were to redo my edit, I would dial back the color and lighten her face a bit, but still leave more than John has in his image. This is my choice, not a criticism of his image. At some point, image editing finally comes down to the "taste" of the editor.

My discussion here is an attempt to discuss alternative approaches to image editing. I am very specifically not trying to point out which image is "better" or "worse" but that there are multiple ways to a goal in Photoshop + ACR. And -- not everyone always even agrees on the final goal.

John's editing moves are described very well. I think of mine as more "intuitive" -- perhaps.....for me anyway. I like John's result quite a bit.

This has been an entertaining discussion, John and Madelaine. I want to thank both of you.

I will try several of John's techniques he described in this thread.

Interesting debate. I'd love to see other spins on the image using different editors, esp. LightZone, my quicker-picker-upper of choice ...

Dive In!

Originally Posted by pathfinder Dive In!

I'm thinking about it ... not sure I can do the program justice tho. I'd hate to turn people off of a good program just because of my poor skills. Any REAL LightZone (or other editor/converter) gurus around?

If no one else replies I'll try to find time this weekend. That's assuming divamum's willing to let me mangle her image in the name of science.

Originally Posted by MarkR That's assuming divamum's willing to let me mangle her image in the name of science.

Have at it..... this poor child's image has now been blasted all over the internet during this experiment, but as long as it doesn't get pulled out of context and turn up somewhere else I'm cool with it!

Originally Posted by divamum Well, I know how to find out what they are by using the dropper and reading the info pane, but I'm still not sure what they're *used* for, if that makes sense. When I see people discussing the numbers I have no idea what they reference.... or why we need those numbers. Are there specific numbers we "should" have for any given colour and against which we "should" measure the colours in our own image? This is the part I'm having difficulty understanding.... (I suspect it's very simple, I just haven't had that "aha" moment yet - I feel a bit like I'm back in 9th grade algebra waiting for the penny to drop!)

PF gave a good summary of what numbers to look for in an area that should be neutral. In the best of cases, by the time you have set the white balance during RAW conversion, you shouldn't have to worry about it in PS. But not everything goes according to the best case, so it is useful to understand how to make white balance corrections later.

Another use of the numbers is for correcting flesh tones, and there are a number of guidelines that are frequently cited. Obviously, there is variation by ethnicity and by individual as well, but Dan Margulis recommends a couple of rules of thumb for Caucasians: in CYMK, yellow should be at a minimum equal to magenta and possibly as high as one third greater, while cyan should range from a fifth to a third of magenta. In LAB, both A and B should be positive, with at least as much B (yellow) as A (magenta), sometimes more. Like all recipes, these should be taken as a starting point, not the final word.

Originally Posted by MarkR I'm thinking about it ... not sure I can do the program justice tho. I'd hate to turn people off of a good program just because of my poor skills. Any REAL LightZone (or other editor/converter) gurus around? If no one else replies I'll try to find time this weekend. That's assuming divamum's willing to let me mangle her image in the name of science.

Not sure that either Jim or I did justice to our approaches either. The main point is to show your steps carefully and in detail. That way, others can learn and possibly make suggestions. My idea in posting this was to make the process and the thinking behind it very concrete.

Originally Posted by rutt Not sure that either Jim or I did justice to our approaches either. The main point is to show your steps carefully and in detail. That way, others can learn and possibly make suggestions. My idea in posting this was to make the process and the thinking behind it very concrete.

I'll give it a shot this weekend. One of LZ's strengths is that you can easily white balance regions of an image independent of others. It will be interesting to see if it works on this image.

Originally Posted by Richard PF gave a good summary of what numbers to look for in an area that should be neutral. In the best of cases, by the time you have set the white balance during RAW conversion, you shouldn't have to worry about it in PS. But not everything goes according to the best case, so it is useful to understand how to make white balance corrections later. Another use of the numbers is for correcting flesh tones, and there are a number of guidelines that are frequently cited. Obviously, there is variation by ethnicity and by individual as well, but Dan Margulis recommends a couple of rules of thumb for Caucasians: in CYMK, yellow should be at a minimum equal to magenta and possibly as high as one third greater, while cyan should range from a fifth to a third of magenta. In LAB, both A and B should be positive, with at least as much B (yellow) as A (magenta), sometimes more. Like all recipes, these should be taken as a starting point, not the final word.

It does take a little experience to learn to use the dropper readings, but I think it's the single most important lesson I've learned about post processing -- don't trust your eyes. This isn't to say that you shouldn't use your eyes; you must. But your eyes will often deceive you into thinking that all is well when it could be better. For example, the slight cold cast in this shot is really no big deal until one starts to enhance color, but then it becomes a real landmine, preventing color enhancement from adding the warmth that this shot needs. Both Jim and I started out by worrying about color balance and by sampling parts of the image to try to figure out if it where it stood, even though we used different tools to achieve the color balance adjustment after we made this decision. But notice that neither of us really focused on the measurements of the flesh when we did this.

I nearly always use LAB measurements to measure, it's very easy and intuitive.

1. A+B positive == warm.
2. A positive, B negative == purple
3. A negative, B positive == green
4. A+B negative == cold.
5. Flesh should be A+B positive, A<=B.
6. Vegetation should be A negative, B positive and B much further from 0 than A, as much as 2x
7. Blue skies should be B very negative and A not negative, pretty close to 0.

As Richard says, don't obey the rules slavishly, lot's of times breaking or bending them is what makes a shot. But knowing them helps know what we got. Here is an example where some flesh has B negative.



There is blue in some of her flesh and it makes it purple. But there is some flesh with that falls inside the guidlines above. That lets us see that the light is blue and that's what's making her flesh look blue where it's in the blue light.

OK, so here was my reasoning when I first examined the picture: The flesh was OK, maybe a little too magenta (A>B) in the forehead, but basically OK. Hmm, what else? Her dress? Should the white parts be neutral? They measured a little warm (A+B positive). Could that be right? I guess it could be. No law that the dress had to be really white. I didn't think about the shoes, but that's not a bad idea. They are way lower than her face, which is what really matters, and the light might not be the same, which might have been why I didn't think about them. What about that wall behind her. It measured distinctly cool (A,B negative). Could this be right? I suppose it could be, but is it what we want? If this shot is to capture the sweet light of late afternoon, the light would be warm and a light colored near neutral area would reflect that and be slightly warm. So I decided to push it just a touch toward warmth (you can see my curves are pretty subtle and the move is small, hard to see before I pumped up the colors much later in the process.)

One advantage of working in RGB curves in PS over white balance in LR is that the curves can apply the adjustment more selectively. In this case the greatest impact was on the quatertones with very little change to the shadows and lightest highlights. This means that I did warm her face, but less than I warmed the wall. This fact accounts for one of the biggest differences between my result and Jim's. Mine has warmth, but less hot flesh. In fact, one of the greatest (current) advantages of PS over LR is it's ability to deal with mixed casts of various kinds. Perhaps the most common of these is when shadows are lit differently than highlights, something that RGB curves handle pretty well.

Hope this helps.

My reasoning with this shot

I am taking the liberty of repeating rutt's listing of LAB values - I think they are that important
1. A+B positive == warm.
2. A positive, B negative == purple
3. A negative, B positive == green
4. A+B negative == cold.
5. Flesh should be A+B positive, A<=B.
6. Vegetation should be A negative, B positive and B much further from 0 than A, as much as 2x
7. Blue skies should be B very negative and A not negative, pretty close to 0.

I would add a few more

8. a and b 0 ( zero ) == a true neutral

9. L less than 5, a very dark tone near black - but color depends on the a and b values

10. L very close to 100 - a VERY bright color, near white - but again depends on the a and b values



When I was editing Madelaine's image, I read several areas for WB as rutt did, in addition to looking at the image on a calibrated monitor.

I did finally measure skin tones in CMYK on her chest wall just anterior to her axilla after choosing my WB point on her shoe, and found them typical Causasian skin tones which helped me think I was on the right track choosing the white balance that I did. I also looked at the color of the shadows in the black wrought iron fencing, the black in the ribbon in her hair, and the black lamp post, but finally chose to use the black shoe as my first choice.

I would have preferred a better, brighter neutral to use, but didn't see any I felt were trustworthy enough to base the image on. I had less faith that the wall was a true neutral, open to the sky as it was.