Help needed on Color Correction..please read

michael972

Hello:

Thanks for reading. I have been struggling with color correction on photographs. The examples were below were shot in JPEG and opened in Photoshop CS. The first is the original without any post. The second image has color cast correction applied to it as described in scott Kelby's book "the photoshop cs book for digital photographers" The technique is the one described on page 118 chapter 5, using the threshold to determine the darkest and lightest areas of the image. My main complaint is that there seems to what I can only describe as a "sheen" on the photograph after I use this technique. It is almost identical to the sheen I see after applying auto levels or auto contrast to an image. Please take a look at the two samples below, and tell me I am not crazy.

Original:



And second, the Kelby technique sample:

digismile

Colour Cast
Hey Micheal,


First of all, I don't think there was very much wrong with the colour on the original photo. Maybe only needed some contrast changes.

The sheen you see is a blue colour cast. I haven't got Scott's book to refer to, but here is how I checked your photo. First of all, my assumption is that the woman's shorts were some kind of shade of white, as was the logo on the man's chest on the right.

I placed a colour checker on the bottom of the shorts and on the man's chest. The RGB values for the shorts were R143,G150,B178. The man's shirt was 176,181,200.

As all neutral colours from white through black have an equal number for R,G, &B, the high number for the BLue channel indicates the blue colour cast.

There are many ways to adjust this. I usually use curves, but have also used levels and colour balance. They all will make this adjustment almost identically.

First, add the colour checker spots that I mentioned.

The, for this photo I added a curves adjustment layer and selected the Red channel first. Go to the very top right corner and drag the corner horizontally and watch the number for the Red channel. It should be increasing. Increase it until it matches the number in the blue channel (in my example, I kept going until it reached 178).

Now select the Green channel and do the same thing, increasing the G value to be the same as the Red and the Blue.

This simple correction gets rid of the colour cast and most of the problems. Generally speaking, when you adjust highlights to neutral, you bring up the lower numbers to match the highest number. If we were adjusting a dark shadow to neutral, we would adjust to the lowest of the 3 numbers for RGB. The advantage of adjusting "to the numbers", is that the numbers don't lie, but your monitor might. If my photo has something that I think should be neutral, I always check it to see if my eye and monitor are correct.

This adjustment has created another issue, ie. it has blown out some highlights which might be undesirable, but can be corrected in other ways.

The following photo simply adjusts the woman's shorts to neutral to show that the sheen is gone.

Brad

With Blue Colour CAst


Curves adjustment Layer to increase Red and Green highlights

wxwax

Excellent post, Digi, thanks for the education.

DavidTO

A lot of Kelby's techniques seem to be geared toward pre-press. I know nothing about it, but I do know this:

The technique he describes with the threshold thing involves setting custom point for the shadows, midtones and highlights eyedroppers.

The shadows are set at 20
mids at 128
and highlights at 240

This leaves 20 points at the bottom of the scale and 15 at the top that go unused. It seems to be a very safe technique, with no risk of losing either shadow or highlight information.

One thing you could do is to set the shadows at a lower point and the highlights at a higher point.

In short, the sheen you're seeing is that the blacks are a bit milky and the highlights don't pop.

The real test would be for you to order prints of those two images and see which one fares better in print. Obviously the clamping down on the range hurts the RGB monitor look of the shot, but maybe the other looks better on print? That's my guess.

A great way to see this illustrated, by the way, is to use the threshold or levels to show a histogram of the image. On the second you'll see the unused bits at either end, and on the original you'll see the full range used.

One great thing about what Kelby suggests is that if you set the RGB values of your shadows, for instance, to 20/20/20, it will make your shadows neutral (remove any color cast) from the shadows (or highlights, depending). That part of his tip is still valid, even if you change 20/20/20 to 5/5/5/.

Also, the point in the picture that you click the midtones (grey) is very important, as subtle differences can cause color shifts. I usually end up "shopping around" by clicking on various points that I believe to be neutral and keep going until I find one that creates a pleasing color balance.

Baldy

DavidTO wrote:

A lot of Kelby's techniques seem to be geared toward pre-press.

I agree with David. It's true that prints cannot have the same snap that a monitor can because the blacks aren't quite as black and the whites not so white. But I don't think that's reason to make your online version a little bit washed out with dark gray blacks.

digismile

DavidTO wrote:

Also, the point in the picture that you click the midtones (grey) is very important, as subtle differences can cause color shifts. I usually end up "shopping around" by clicking on various points that I believe to be neutral and keep going until I find one that creates a pleasing color balance.

I think that this is probably the most important point in understanding what the shadow, midtones, and highlight samplers are actually doing.

They literally force the selected colour value to be equal RGB values, ie something like 20,20,20 for the shadow point, 240,240,240 for the white point, etc. So, in making these changes, we change all values in that colour channel and similar tonal range. Therefore, if we force a none-neutral point to be neutral (instead of its natural colour), we have likely introduced some kind of colour cast. It's especially noticable when you use the midtone sampler as David indicated.

Just to see this point in action, use the photo in this thread, put a colour sampler on something like the grass to see what it's RGB values are. THen add a curves adjustment layer and click on the grass sample point with the midtone sampler. Look at the new values. Voila, neutral grass ... but very strange colour shift.

Brad

DavidTO

Just to illustrate the point about what's going on with the levels, I pulled your images down from my browser, opened them in PS and got levels. Attached are the two levels windows, the top being the original and the bottom being the one with the "Kelby Conversion".

michael972

Thank you for all the fantastic informatoin.

DavidTO

Another interesting thing about your two images.

I converted them both to CMYK (printing color space) and there was no noticeable shift on the Kelby Conversion. The other did shift and became more like the Kelby. So, for me, that confirms that using his threshold method will produce more accurate results if you're printing.

Argusphoto

3 people
I have another suggestion on correcting this picture. I went to Levels and adjusted the mid range to 1.65 and then Image>Adjust>brightnes/contrast and increased the contrast to +5

lynnma

Argusphoto wrote:

I have another suggestion on correcting this picture. I went to Levels and adjusted the mid range to 1.65 and then Image>Adjust>brightnes/contrast and increased the contrast to +5

Fabulous tutorial everyone, thanks so much I'm printing this out for study, very very informative.
Lynn:D