Flower girls

Aviator327 · November 1, 2005

Took this pic last year at a friends wedding. Playing around with PSCS. Don't quite know how to get the skin tones a little better. Used my 17-40L for this shot.

51661397.IMG_2799after.jpg

rutt · November 2, 2005

Wow, what happened here? Beside the blown spots and magenta cast on the faces, how did the roses in their hair become dead looking? Yet, the background colors are all very plausible. Do you know? Or maybe someone can take a guess; I certainly don't know how it happened.

Yet it is very pertinent to stuff I've been working on. I'm supposed to write a summary of Chapter 8 of Dan Margulis' LAB book and this shot is a good place to practice the technique he showcases there.

The worst problem is that the faces are blown in places (as light as they can be) and are much more magenta than yellow, which almost always looks wrong. Truly global moves, such as LAB curves could change the color balance but the background is fine and we don't really want to move the white balance except in those blown skin patches.

I didn't worry about the roses yet, only the faces. You can always address the roses with a selection if you want to.

I managed to get this result:

Here are the steps:

Convert to LAB
Make a duplicate layer
Use the color sampler to choose a reasonable color from one of their necks.
With the duplicate layer selected, paint some big splotches to cover their flesh:
Use LAB curves on the duplicate layer to fine tune the color balance and saturation of the brush strokes so they are nice healty flesh tones, more yellow than magenta and plenty of both.
Set the blending mode of the duplicate layer to Color. This means that only the color comes from the duplicate layer, but all the luminosity and detail come from the original.
Bring up the blending options for the duplicate layer by right-clicking on it (I think control-click on the mac).
Now I needed to limit the blend, so that the color doesn't spill over onto the dresses or hair or teeth. I did this with the Blend If sliders.
First I limited to blend to exclude the darkest parts of the image:
.
Note the split slider. This defines a range of blending. The right endpoint defines the end of the range. Colors lighter than this are blended, i.e., come from the duplicate layer. The left endpoint defines the start of the range. Colors darker than this are not blended, i.e., come from the background layer. In between, there is more blend as the colors lighten. Split the slider with option-click (Mac).
This excluded a lot of what I didn't want from the blend, but there were still places on the dresses which were within the brightness range and were undesirabley blended. So I also adjusted the blend if settings for the A channel.

Here I excludes colors with some green. Again note the split slider to make the transition areas look more natural.
Here is the image at this point:

Looks better, but I found one bad side effect. I'd like those lips to be redder, instead of flesh toned. At this point, I gave into temptation and used a layer mask on the duplicate layer (with the paint strokes) to reveal the color of the lips from the original.

The theory behnd some of this is a little intense and I'm planning to discuss it when I write the chapter summary. Suffice to say, this wouldn't have worked in RGB, or at least not nearly as well. The color blend is a really great discovery.

I'll bet there are other different ways to do this; perhaps some will yield better results. When you have a hammer, everything looks like a nail.

DavidTO · November 2, 2005

Looks much better Rutt. Chalk this up as one of your successes. Great job.

DavidTO · November 2, 2005

I want to point out here one of the basic things about RGB and its limitations. By definition, the darker the shadows and the brighter the highlights, the more they go towards neutral.

This is because white is 255R,255G,255B. The brighter a color becomes, the more it moves towards neutral white. That is a big part of the reason why you lost so much color on those faces. RGB cannot describe a very bright fleshtone.

rutt · November 2, 2005

DavidTO wrote:

I want to point out here one of the basic things about RGB and its limitations. By definition, the darker the shadows and the brighter the highlights, the more they go towards neutral.

This is because white is 255R,255G,255B. The brighter a color becomes, the more it moves towards neutral white. That is a big part of the reason why you lost so much color on those faces. RGB cannot describe a very bright fleshtone.

That explains the blown sections on the faces but not the dead roses. Also why did yellow give up the ghost before magenta?

edgework · November 2, 2005

The background colors may be plausible but the wooden structure is probably close to neutral and at the moment it, like the faces, is woefully short on yellow. I used the beams as a reference in lab to pull them to neutral, did some quick retouch on the faces and took the same color mode painting approach for the blown out areas. The roses needed help on their own. Back in rgb I took some cyan out of the greens with selective color.

Aviator327 · November 2, 2005

rutt, Thanks for the detailed info. The only adjustments I have used so far in PS are Levels, Curves, contrast and saturation. I haven't even begun to work with duplicate layers yet. I have some books on PS7 but none for pscs. I am going to have to experiment with layers and see what I come up with. As for the flowers, they are artifical. I am learning something new every day. Thanks.

Aviator327 · November 2, 2005

edgework wrote:

The background colors may be plausible but the wooden structure is probably close to neutral and at the moment it, like the faces, is woefully short on yellow. I used the beams as a reference in lab to pull them to neutral, did some quick retouch on the faces and took the same color mode painting approach for the blown out areas. The roses needed help on their own. Back in rgb I took some cyan out of the greens with selective color.

Very nice retouch. I think I was messing around with Selective color and may have moved the cyan slider to much. I am not even going to ask how you did the flowers. I need to get me a book on pscs.

:

rutt · November 2, 2005

edgework wrote:

The background colors may be plausible but the wooden structure is probably close to neutral and at the moment it, like the faces, is woefully short on yellow. I used the beams as a reference in lab to pull them to neutral, did some quick retouch on the faces and took the same color mode painting approach for the blown out areas. The roses needed help on their own. Back in rgb I took some cyan out of the greens with selective color.

Yeah, that looks very nice. As I said, when you have a hammer, everything looks like a nail.

edgework · November 2, 2005

Aviator327 wrote:

Very nice retouch. I think I was messing around with Selective color and may have moved the cyan slider to much. I am not even going to ask how you did the flowers. I need to get me a book on pscs. :

The roses were straight out of Dan's book. Bottom anchors in a and b channels pulled 20 points or so to the right, with a mask to isolate the petals. Took about 20 seconds, including mask.

Even though Dan Margulis keeps coming up with new innovations, something he said back when his Professional Photoshop first came out is so utterly simple it's easy to forget: Whatever's wrong in one part of the image is usually going to be wrong in the entire image. In this picture the roses were the odd exception, but the pink faces were symptomatic of an overall yellow deficiency.

rutt · November 2, 2005

edgework wrote:

The roses were straight out of Dan's book. Bottom anchors in a and b channels pulled 20 points or so to the right, with a mask to isolate the petals. Took about 20 seconds, including mask.

Even though Dan Margulis keeps coming up with new innovations, something he said back when his Professional Photoshop first came out is so utterly simple it's easy to forget: Whatever's wrong in one part of the image is usually going to be wrong in the entire image. In this picture the roses were the odd exception, but the pink faces were symptomatic of an overall yellow deficiency.

OK, so you've won a valuable prize. You get to write a chapter summary for the LAB book discussion group.

pathfinder · November 2, 2005

edgework wrote:

The roses were straight out of Dan's book. Bottom anchors in a and b channels pulled 20 points or so to the right, with a mask to isolate the petals. Took about 20 seconds, including mask.

Even though Dan Margulis keeps coming up with new innovations, something he said back when his Professional Photoshop first came out is so utterly simple it's easy to forget: Whatever's wrong in one part of the image is usually going to be wrong in the entire image. In this picture the roses were the odd exception, but the pink faces were symptomatic of an overall yellow deficiency.

Nice retouch job, edgework.

The first time I read that statement from Margulis years ago, I had real trouble buying it - but the more I work in PS, the more I 've come to agree with it.
Congratulations - Now rutt will get you to sharing the work with all of us on the LAB discussion group, so don't let us down!!

:

rutt · November 2, 2005

pathfinder wrote:

Nice retouch job, edgework.

The first time I read that statement from Margulis years ago, I had real trouble buying it - but the more I work in PS, the more I 've come to agree with it.
Congratulations - Now rutt will get you to sharing the work with all of us on the LAB discussion group, so don't let us down!! :

Yeah, I mostly don't forget this and I'm a little embarrassed that I forgot it this time. I was just hyptomized by the color blend technique I was in the process of writing up for the discussion group.

An amazing thing about getting good at all this. Once you get there, you can make almost any color happen almost anywhere in the image. Each new giood technique adds to that ability. Then you have to learn restraint and you have to remember the basics, such as that quote.

mereimage · November 2, 2005

thanks for the lesson Rutt.
The overall color has been improved but there is a 'greyness'to their cheeks and forehead that seems unnatural... has the look of excessive noise suppression or use of the cloning stamp. Its there in the orig. image at the start too

...................Mereimage....Or perhaps its makeup?

edgework · November 3, 2005

rutt wrote:

OK, so you've won a valuable prize. You get to write a chapter summary for the LAB book discussion group.

Bring it on!

edgework · November 3, 2005

rutt wrote:

Yeah, I mostly don't forget this and I'm a little embarrassed that I forgot it this time. I was just hyptomized by the color blend technique I was in the process of writing up for the discussion group.

Same here. When I first read the LAB book, it was as if everything I'd been doing in CMYK and RGB no longer existed. Which lead, as is the case in Dan's book, to some convoluted efforts to achieve what I'd ordinarily been able to approximately equal with much simpler approaches. But the blending sliders, in particular, are an example of something I'd ignored for years, and once I "got it", I've found ways to integrate them into my work flow everywhere I look. It's kind of a three-step approach: 1) figuring out what the heck he's talking about; 2) figuring out how to use it in a given job; 3) figuring out how it fits in with what you already know. Takes a while.

Aviator327 · November 3, 2005

mereimage wrote:

The overall color has been improved but there is a 'greyness'to their cheeks and forehead that seems unnatural... has the look of excessive noise suppression or use of the cloning stamp. Its there in the orig. image at the start too...................Mereimage....Or perhaps its makeup?

The girls were not wearing makeup. The image I posted was me messing around with PS. I will post the origional image (no processing) later on. Exif data Shot at 1/180sec, f5.6 ISO-200

rutt · November 3, 2005

Aviator327 wrote:

The girls were not wearing makeup. The image I posted was me messing around with PS. I will post the origional image (no processing) later on. Exif data Shot at 1/180sec, f5.6 ISO-200

If you post the original and allow an original sized download, I'm sure we'll be all over it and you'll end up with at least one (and maybe three) nice retouches.

rutt · November 3, 2005

How to get a full sized original?

Aviator327 · November 3, 2005

Here is the full size original. May need some of your expert advice.:):

Here.

Moderator Edit: I changed the inline image to a link.

rutt · November 3, 2005

Aviator327 wrote:

Here is the full size original. May need some of your expert advice.:):

Here.

Moderator Edit: I changed the inline image to a link.

This is not the highest resolution image that a 10D can shoot. This image is 1536x1024 but the camera can shoot 3072x2048; however, if you shoot in something less than L or RAW, you might have gotten this. Can you please check, what is the resolution of your original file? You can do this in photoshop if you have it by Image->Image Size and it will tell you. Don't actually change anything, though (IMPORTANT.)

Aviator327 · November 3, 2005

rutt wrote:

This is not the highest resolution image that a 10D can shoot. This image is 1536x1024 but the camera can shoot 3072x2048; however, if you shoot in something less than L or RAW, you might have gotten this. Can you please check, what is the resolution of your original file? You can do this in photoshop if you have it by Image->Image Size and it will tell you. Don't actually change anything, though (IMPORTANT.)

Rutt, That is the correct size. I had the 10D set for Small (fine). I have only shot a few images in RAW. I shoot small most of the time. Thanks.

Flower girls

Comments