Wedding Shots/Color Correction

ridetwistyroads

The sky's the limit here folks. I have thick skin! :wink

A close friend of mine got married a little while ago, and I just worked up the courage to ask the collective here for some critique. Her hired photog fell through, so I stepped in. I was going to shoot anyways, for practice kicks and giggles, but being under that much pressure had me quite scared.

I will add, she's happy with the shots, thrilled, acually. But, I throw myself upon the whipping post: I strive to improve!

Thanks all,
Chris


edit; gallery link to a few more

























All shot with D70, natural (stage/house) light. I used two lenses, 24-85 f3.5-4.5 for the poses, and a 70-300m f4-5.6 for the ceremony. The long lens is deffinately a budget lens, and, I certianly would have liked to have had a faster lens for all around use (hence the hi iso noise). I have tried to process some noise out, but am fairly new to the world of post.

Again; Be harsh!! :uhoh
Thanks!

Andy

i moved your thread. the rulez of the whipping post forum are *one* image for critique. you can still get critique here in the people forum.

you may submit one image per week to the whipping post.

please check the rules thanks!

leebase

The color seems off. Did you shoot raw? Did you use a custom color balance for the house lights?

The photo of the b&g with the kids could use some straightening.

But hey! You took some great photos that I'm sure the b&g will cherrish. There's a big difference between what photographers will tell you about your photos and what the people in them will tell.

They won't be nit picky

Lee

ian408

Twisty,

Compositionally, they're very nice. But each suffer from white balance
problems. Shooting in low light is difficult and creates unique shadowing
challenges. #2 demonstrates the difficulty. The groom's face is in shadow
while the others are not. I would like to suggest a little fill flash might have
helped though the venue may not have allowed for that.

I'll leave the real critique to those better able to judge wedding shots.

I think you did well given the circumstances.

Ian

rutt

These are really nice wedding shots and well worth saving. The color balance is a real problem, but it can be fixed pretty easily. If you shot in raw, it's particularly easy to fix. Use the color balance tool in ACR to balance on the wite of the wedding dress.

If you didn't shoot in raw, you can use curves to fix. The white wedding dress will be a godsend. You know it should really be white. Try converting to LAB and moving the A+B endpoints inward until spots on the dress measure A=0, B=0.

I did this with my favorite (and the easiest one of these).

Here is the result:

Here is what I did.
1. Mode->LAB

2. Set a color sampler point on a strategic point of the white dress:


3. Image->Adjustments->Curves. Here are my A+B curves:


I knew how to write these curves because I looked at the Info palette while I did it:



Point 1 shows that the initial values were A=6, B=18. These values add up to a color that is really not white, very warm in fact. One of the beauties of LAB is that A=0, B=0 is neutral independent of brightness (lumonosity). So in this shot the dress, groom's color, and even his tuxedo, should all be approximately A=0,B=0. But the lighter the spot, the more visible will be variations from neutral. So I moved the warm endpoints of the A and B curves to flatten out the warmth and bring the selected point to neutral.

I've found this technique to be invaluable in all kinds of situations, especially jpegs shot under tungsten.

rutt

I want to add one more thing. Notice how the rest of the shot improved "automatically" when I balanced the white on the bride's dress. This is typical of this kind of color correction. If you can find a point on the picture with a color you know for sure and use global methods (curves preferably) to make the image match that color, often the rest of the shot will also snap into place.

lynnma

rutt wrote:

I want to add one more thing. Notice how the rest of the shot improved "automatically" when I balanced the white on the bride's dress. This is typical of this kind of color correction. If you can find a point on the picture with a color you know for sure and use global methods (curves preferably) to make the image match that color, often the rest of the shot will also snap into place.

rutt.. you are a genius.. I just tried it and it worked like a charm.. could it be finally sinking in for me???? could it?????

Thank you soooo much...

lynnma

lynnma wrote:

rutt.. you are a genius.. I just tried it and it worked like a charm.. could it be finally sinking in for me???? could it?????

Thank you soooo much...

ok.. so what do I do with my shots who are not wearing a white dress...:uhoh

Dee

Can you do this without converting to LAB?

rutt wrote:

If you didn't shoot in raw, you can use curves to fix. The white wedding dress will be a godsend. You know it should really be white. Try converting to LAB and moving the A+B endpoints inward until spots on the dress measure A=0, B=0.

Color balance is the area I'm trying to work on now. There are so many different ways of accomplishing the same goal in photoshop, that's why I asked, out of curiosity.

lynnma

so just to prove out Rutts point again (I'm so excited) I found this terrible shot (ignore it) but he has a WHITE shirt.. so I experimented.. heres the result..


after lab curve making a and b zero



I mean it's not perfect by any menas but it's a lot better color don't ya think?

Rufio220

I like your pictures. They are nice, I like them.

My friend does pictures and movies of wedding for money. It pays good if you need another job.

rutt

The subject of Dee's post was "Can you do this without converting to LAB?" The answer is, "Yes, but why would you?" For these particular shots, converting to LAB is a magic bullet. They need a radical color conversion. They have an obvious color neutral area. In LAB, you just move the endpoints to zero the A&B values there and you've done at least 90% of the work.

The althernative is to do this in RGB or CMYK. In those color spaces, you have to worry about the brightness of the points you are correcting at the same time as you worry about the color. This makes it much harder (though still very possible.)

What is lost by converting to LAB. In fact nothing. LAB has the widest gamut of any colorspace. RGB->LAB, unlike RGB->CMYK, loses nothing.

So go ahead and torture yourself, but an initial quick fix in LAB is by far the easiest approach here (and in many shots that need major color correction.)

rutt

lynnma wrote:

ok.. so what do I do with my shots who are not wearing a white dress...:uhoh

Dan Margulis compares it to the game of bridge. You have to gather up all the clues you have and put together a picture. If there is a a spot that you know should be white, that's a huge clue, probably the most important one you can have. But you also know that tuxedos are black, for example. Stop lights are red. Sky might be blue, but it can't be green. The ocean can only be red under very special circumstances. Black dogs are black, not blue or cyan. And flesh is some shade of red made up of at least as much yellow as magenta and maybe 30% less cyan (depending on lighting and the person's complexion.)

So, let's take your example. Good, you've neutralized the white spot and accomplished about 90% of the job. But as you noted, it's still not perfect. Why? Use the color sampler on the man's face. It's still too magenta vs yellow. I placed a color sampler point on his forhead:



And read a value of C=5, M=55, Y=32, K=0. OK, here is an easy rule of thumb, except for sunburn, flesh always has at least as much yellow as magenta. So we know this is still not right.

A second LAB curve adjustment can put it right in a hurry. We just want less magenta in the face and want to keep the shirt wihite, right? OK, we'll just pull the magenta side of the curve toward neutral having locked down the neutral point and the green side.

You can place a point on the curve by alt-clicking the image. So I locked down the white point on the shirt by alt-clicking it. I locked down one other point as well, way up in the green part of the curve that I don't want to change. Then I clicked on the pont on the forhead to see where on the curve that was and pulled it toward neutral until I got an acceptable reading from the color sampler at that point: C=22, M=49, Y=51, K=2.

Here is the A curve I wrote:



And here is the image after this second adjustment:



The shirt is still white, but the sunburn is gone.

Although it's very elegant to make your entire correction with one application of curves, often it isn't the easiest approach. With this image, a two stage approach worked well and was very easy. Stage 1: make the radical move to neutralize a known color neutral point. This does the heavy lifting. Stage 2: fine tune to get other important colors right, most notably flesh. Once the first radical correction is done, this is often relatively easy.

ziggy53

rutt wrote:

...OK, we'll just pull the magenta side of the curve toward neutral having locked down the neutral point and the green side.

You can place a point on the curve by alt-clicking the image. So I locked down the white point on the shirt by alt-clicking it. I locked down one other point as well, way up in the green part of the curve that I don't want to change. Then I clicked on the pont on the forhead to see where on the curve that was and pulled it toward neutral until I got an acceptable reading from the color sampler at that point: C=22, M=49, Y=51, K=2.

Here is the A curve I wrote:



... this is often relatively easy.

Rutt,

I am still not clear about your exact procedure on the second correction, relating to "locking" a point and multiple samples.

Can you please say it a bit differently so that it has a chance getting through my thick skull?

Thanks,

ziggy53

rutt

ziggy53 wrote:

Rutt,

I am still not clear about your exact procedure on the second correction, relating to "locking" a point and multiple samples.

Can you please say it a bit differently so that it has a chance getting through my thick skull?

Thanks,

ziggy53

Look at the points on this curve. All but one are identity points; their input and output is the same. At that point in the curve there is no change. Putting a couple of these point on the right (green) side of the curve, locks it down so it won't move when I add a point to the magenta (left) side of the curve and move it. In particuar I have a point at 0,0 and one at 75,75 that preserve the green side of of the curve. Then I was able to point at the point on the man's forehead (with the curve dialog still active) and click. This showed the exact point on the curve for this point on the picture. Alt-Click (Apple-Click on mac) actually adds a point to the curve in the spot as well as just displaying it.

Learning through doing probably works best here. Download Lynn's after image and try writing A curve above. Experiment freely. Nothing will explode.