Color Problems

imann08 · #1 03-06-2007, 10:28 PM

This picture is of me and a girl who is now 19. I was 5 at the time.

Anyways, I don't like what I have come up with here. I am still having problems with the amount of cyan in the skintones. I was having a lot of problems getting the overall darkness down and I eventually used a false RGB profile with a lower gamma to accomplish that. I also think I blurred it too much to begin with but as you can see it needed some work with all that noise. If anyone wants to give this a go then I'd love to see your results and a description of what you did to get rid of the cyan without making the faces look beet red.

DCobb · #2 03-06-2007, 10:46 PM

I don't know if my picture on the right is an improvement or not. Started with Digital Roc and then did a curves and Shad/High. adjustment.
dc

imann08 · #3 03-06-2007, 11:39 PM

There really isn't any difference between mine and yours. I could have used S/H but the entire image needed lightening and not just the shadows. When that is the case, S/H is not the best method. For the noise, I have neat image which I used. How did your skintones turn out? Is there still too much cyan in it?

chillin · #4 03-07-2007, 12:37 AM

a lot of posibilities with this picture...

Gary Richardson · #5 03-07-2007, 02:06 AM

Haven't really tried to deal with the noise, just worked on the colour problems.

As the lighting across this picture is not constant, I dealt with different areas differently.

First did a general levels adjustment on LHS, leaving the wall on the RHS de-selected (masked). Purpose to even out the lighting, the RH wall area is too dominant if a whole picture adjustment is made.

Then used curves to colour correct whole image.

Lastly did a colour adjustment on skin tones only.

imann08 · #6 03-07-2007, 02:31 AM

Gary and Chillin. I like your edits a lot. My main problem is that I have trouble when there is a significant amount of cyan in the skintones as there was here. In my attempts to correct this I just ended up with very red faces that looked worse. I'm sure I'm making a stupid mistake here but I'd love to know what I'm doing wrong. Did you guys get good skintone values? I'm off to bed now but I'll check them in the morning myself. If you could go over the process a little more in depth I'd appreciate it.

Gary, I really liked the idea of yours to separate the two sides. I'd like to think that there is a better way to accomplish this but I don't know without giving it some thought. I actually colored that RHS wall yellow in LAB using the dodge tool in the B channel. Very mildly of course but I don't think it was supposed to be white like the shirts were. Getting that back wall more in line with the side one was important though. I also like that you got more blue out of the little girl's eyes than I did by far. Did you use some painting tool or did it just work out with the curves you used?

Thanks again to both of you.

BillFrey · #7 03-07-2007, 06:06 AM

Hi imann08,

Surprisingly, auto color did a nice job with this one, followed by shadows/highlights, levels with a mask for the near wall, light noise removal, light sharpening.

Bill

Syd · #8 03-07-2007, 09:20 AM

First I straightened it, then I cropped it. That sign on the wall might have been of interest but seeing as part of it was missing I cropped it out and with it went that glaring white wall. The ceiling was of no particular interest either. Auto Color does the trick very well here which I followed with a curve to lighten. I used the Color Replacement tool on that back wall, ran it through Neat Image (noise reduction set to 50%) and sharpened it slightly.

Syd

Gary Richardson · #9 03-07-2007, 09:47 AM

Hi imann08,

What I did when colour correcting is what I usually try first. ie I used a Curves adjustment layer as follows (it works in most cases).

1. Select Color Sample Tool and place a sampler in a Mid-Grey, White (or light grey) and Black (or Dark Grey) areas. In this picture I chose the far wall on the left, a bright area on the RH wall, and the writing on the poster.

2. Create a Curves Adjustment layer.

3. Now I look at the RGB figures on the sampler for the Mid-Grey sample. Select the mid value from the values present. Lets say for example the Red Channel had a value of 150, the Green 140, and the blue 160, then I would be trying to get the values of all channels to be 150 (the mid value). To do this, change the Channel in the Curves dialogue box to Green. Now move your cursor over the spot in your picture where your colour sampler is for Mid-Grey, when over it CTRL+Click and a spot will appear on the Green curve. Click on this and move up with the up arrow key on your keyboard. While you're doing this, watch the figures in your colour sampler box, keep moving until the Green value is the same as the Mid (Red) Value. Now change channel in Curves box to Blue and repeat, this time trying to bring the Blue value down. Once all 3 channels have the same value then you're finished with that sample, move on to the White and Black.

4. Repeat for the White (Light-Grey) and Black (Dark-Grey) areas. For the Whites try to get them all to equal the Highest Channel Value, and for the Blacks try to get them all to equal the Lowest Channel Value.

5. You'll find changing one (Mid, White, Black) "pulls" the others a touch, so it's necessary to re-adjust until you get the nearest set of figures which are equal (or thereabouts) across all channels for Mid-Grey, White and Black. NOTE: The Mid-Grey is the most important to have right, the other two are of lesser importance.

6. Exit out of Curves.

This all sounds very complicated and time consuming, but it is much easier to do than to explain and only takes a few minutes.

For the faces I just selected the skin areas with Quick Mask, converted to a selection, then did a Colour Balance Adjustment layer and increased the yellow and cyan a touch (I don't work with figures, I just eyeball things).

The eyes just came that way with the Curves adjustment.

Hope this helps.

Neillo · #10 03-07-2007, 04:28 PM

Syd,

Like what you've achieved there bud. You've enhanced the detail, reduced the noise and brought back the colours. Good job.

N.

Syd · #11 03-07-2007, 05:34 PM

Thanks Neillo.

Syd

unimatrix001 · #12 03-07-2007, 05:49 PM

as BillFrey mentioned auto color did a really nice job with the colors.
auto color
new layer
desaturate
invert
gaussian blur
blending mode to soft light
ran through noiseware professional
and came up with this

imann08 · #13 03-07-2007, 09:12 PM

Quote:

Originally Posted by Gary Richardson

Hi imann08,

What I did when colour correcting is what I usually try first. ie I used a Curves adjustment layer as follows (it works in most cases).

1. Select Color Sample Tool and place a sampler in a Mid-Grey, White (or light grey) and Black (or Dark Grey) areas. In this picture I chose the far wall on the left, a bright area on the RH wall, and the writing on the poster.

2. Create a Curves Adjustment layer.

3. Now I look at the RGB figures on the sampler for the Mid-Grey sample. Select the mid value from the values present. Lets say for example the Red Channel had a value of 150, the Green 140, and the blue 160, then I would be trying to get the values of all channels to be 150 (the mid value). To do this, change the Channel in the Curves dialogue box to Green. Now move your cursor over the spot in your picture where your colour sampler is for Mid-Grey, when over it CTRL+Click and a spot will appear on the Green curve. Click on this and move up with the up arrow key on your keyboard. While you're doing this, watch the figures in your colour sampler box, keep moving until the Green value is the same as the Mid (Red) Value. Now change channel in Curves box to Blue and repeat, this time trying to bring the Blue value down. Once all 3 channels have the same value then you're finished with that sample, move on to the White and Black.

4. Repeat for the White (Light-Grey) and Black (Dark-Grey) areas. For the Whites try to get them all to equal the Highest Channel Value, and for the Blacks try to get them all to equal the Lowest Channel Value.

5. You'll find changing one (Mid, White, Black) "pulls" the others a touch, so it's necessary to re-adjust until you get the nearest set of figures which are equal (or thereabouts) across all channels for Mid-Grey, White and Black. NOTE: The Mid-Grey is the most important to have right, the other two are of lesser importance.

6. Exit out of Curves.

This all sounds very complicated and time consuming, but it is much easier to do than to explain and only takes a few minutes.

For the faces I just selected the skin areas with Quick Mask, converted to a selection, then did a Colour Balance Adjustment layer and increased the yellow and cyan a touch (I don't work with figures, I just eyeball things).

The eyes just came that way with the Curves adjustment.

Hope this helps.

I appreciate your input. In the future, you don't have to be a descriptive for the more simple stuff like setting neutrals. Despite the way I come across, I actually know what I'm doing for the most part. No need to describe how to use any tools or anything.

I do have a couple of questions as to the way you go about things though. First, you mention that the midtone grays are the most important to get right colorwise. I have heard that this is not the case. As human eyesight goes, color differences are much more noticeable in highlights than in shadows and contrast is much more noticeable in shadows than in highlights. Therefore, if you want to get a color right then you should make sure your highlights are good. Your opinion?

Second, I can't imagine going on monitor appearance alone. I go by the numbers first and foremost and use the monitor when I am within the proper range of color for the image. This is for color, highlights, and shadows. Other things, such as USM, I will use the monitor almost specifically. This is the main reason that I posted this image. I was off in the numbers. Drastically in some cases. It looked alright though so I wasn't that upset with my final.

I am a loyal follower of Dan Margulis and therefore don't consider masks sometimes when I should and never use anything with the word auto in it. In this image the Auto Color Correction works because there aren't any catchlights or other things that could seriously ruin it.

So, basically, I'm not having a problem with the different ways to do things or how to reduce the noise etc. I am simply getting whacked out cyan values in the skintone. I see that you used a mask to attack the skin separately from the rest of the image. That may be the only way of doing it but I like to exhaust my options before going to masks.

Gary Richardson · #14 03-08-2007, 01:23 AM

The trouble with using Highlights to determine the colour balance of an image is that they can be slightly blown out if the lighting is not quite correct and therefore give unreliable results.

Also although the eye is more sensitive to highlights, the camera used may not be, CCDs (or whatever sensor your camera uses) are most sensitive and reliable in the mid-range sector of their response curve, tending to drop off towards the lighter and darker ends.

But mostly my comment about Mid-Grey being best is based on my experiences. I've often corrected based on Highlights only, only to find the result is way out from what it should have been.

I appreciate why you like to work by numbers, but I just don't. I'm pretty familiar with my monitor, and trust my eye, so have a good idea of whether things will turn out right or not.

Not overly scientific I know, but then I'm not trying to do this for a business, for me it's just an enjoyable pastime.

As for masks, I just use them all the time. There are so few occasions where a global adjustment gives you the result you want, and I like the control and fine adjustment that masks allow.

imann08 · #15 03-08-2007, 10:53 PM

Quote:

Originally Posted by Gary Richardson

The trouble with using Highlights to determine the colour balance of an image is that they can be slightly blown out if the lighting is not quite correct and therefore give unreliable results.

Also although the eye is more sensitive to highlights, the camera used may not be, CCDs (or whatever sensor your camera uses) are most sensitive and reliable in the mid-range sector of their response curve, tending to drop off towards the lighter and darker ends.

But mostly my comment about Mid-Grey being best is based on my experiences. I've often corrected based on Highlights only, only to find the result is way out from what it should have been.

I appreciate why you like to work by numbers, but I just don't. I'm pretty familiar with my monitor, and trust my eye, so have a good idea of whether things will turn out right or not.

Not overly scientific I know, but then I'm not trying to do this for a business, for me it's just an enjoyable pastime.

As for masks, I just use them all the time. There are so few occasions where a global adjustment gives you the result you want, and I like the control and fine adjustment that masks allow.

I'm enjoying this conversation. With highlights, I wasn't talking about ones that are blown out. This would be determined by the info palette obviously. As far as midtones being favored by the CCD, that is pretty much irrelevant as far as I know. If we are more sensitive to highlights than to midtones and shadows then it really doesn't matter what the CCD was favoring when it captured the shot. The human eye is more sensitive to highlights regardless and so therefore, they have to be right on whereas shadows can be off a tad without a problem. In the case of this photo, it was a print from film and not a digital camera shot so this wouldn't even apply here at all.

As for personal experiences, that's up to you. If you can trust your monitor and your eyesight enough then more power to you. I can't or at least, I wont.

As far as masks go, I would expect most images taken by digital cameras are going to have global issues. Now, they may not be uniform casts or whatever but they will be global. It is when you get into film prints of years past that you get into issues that are not global and that is due to reasons other than what the camera did. It's how the paper and print have held up over time. Understand that I am not making an argument against masks. They are extremely valuable and useful as is the case here. I just have one thing with making general corrections. I will give curves a go first. Sometimes masks become apparent as they did here. Most of the time masks will be apparent from the minute you look at the photo though as is the case when stains are an issue or you want to artificially enhance an area in a way that never existed such as enhancing the blue of skies, the darkness of fleshtones, etc.

Please understand that this is a debate of sorts and not an argument on my behalf and would love you to prove me wrong if that is the case. You definitely have more real life experience than I as my knowledge comes more from books than actual photos. I consider this a learning experience for me. Thanks.

Neillo · #16 03-09-2007, 02:24 AM

imann08,

No doubt we all understand and acknowledge that you know the basics and that you would prefer a less 'descriptive' answer without the need for highlighting the most basic of commands and actions.

However, please also bear in mind that some of us may not neccessarily have used these actions before in order to rectify our own colour problems—so these fuller, more rounded answers by the experts among us are a VERY helpful source of knowledge and expertise.

We should encourage fuller answers, not everyone is to your standard.

regards,
N.

Gary Richardson · #17 03-09-2007, 02:42 AM

Quote:

As far as midtones being favored by the CCD, that is pretty much irrelevant as far as I know. If we are more sensitive to highlights than to midtones and shadows then it really doesn't matter what the CCD was favoring when it captured the shot.

But its the very data that that CCD produces that creates your picture, and is therefore the data that you are now manipulating in hope of getting things to look how you want. If the data is not consistant across the lighting of the scene (as will be the case if the light and dark areas are outside the linear response curve of the device), then you cannot "trust" the data for the light and dark areas to be accurate, and therefore cannot "balance them" to give a true colour corrected picture, the mid tone areas are the ones which will give the more reliable results.

This is of course a very much simplified explanation, as other factors will contribute to things. For example, the individual resposes of the R, G, B, elements of the sensor will also have some effect, the response curves for each colour are likely to be different, and will have been set up to be most "in step" with each other in the mid-tone frequency range.

If you're using numbers to balance things it is your eye response that is irrelevent (it would be all important if you were judging things by eye).

But as you are working with numbers, you want to work with the "best" numbers (ie the ones that are most reliable), and these are the ones produced when your camera's sensor is in its optimum response zone (ie mid-tone).

As for digital images needing only global changes, that is just not the case. As I've already described, sensor response is not the same for all colours in all light conditions. Further to that, sensors are more reliable in the central viewfinder zone, and trail of towards the edges, where the light is not hitting them at the same angle as it is in the centre (the focal distance is slightly longer towards the edges of the sensor).

Whilst this is almost insignificant in most normal lighting scenarios, it becomes more significant in poor lighting, and its these situations where more localised adjustments are usually necessary. It is also obviously affected by the focal length of the lens being used.

albatrosss · #18 03-09-2007, 10:57 AM

Swampy has a nice tutorial on this at:

http://home.earthlink.net/~lpswampy/

Works nicely with your photo.

sunfly · #19 03-09-2007, 01:48 PM

1) Separate layers for different elements

2) Levels

3) Curves

Cropping, cloning, dodging, neat image, and a final noise layer.

Thanks to all experts for all their tips. :-)

Sherry

imann08 · #20 03-09-2007, 05:43 PM

Quote:

Originally Posted by Neillo

imann08,

No doubt we all understand and acknowledge that you know the basics and that you would prefer a less 'descriptive' answer without the need for highlighting the most basic of commands and actions.

However, please also bear in mind that some of us may not neccessarily have used these actions before in order to rectify our own colour problems—so these fuller, more rounded answers by the experts among us are a VERY helpful source of knowledge and expertise.

We should encourage fuller answers, not everyone is to your standard.

regards,
N.

I was not trying to brag or anything like that. I will actually freely admit that many of the people on this board are much much better than I am. The only reason I said what I said is because I didn't want Gary or you or anyone else to have to take more time than was necessary to explain what they did. As with everyone, there are many times where I need drawn out explanations but if I can save someone a little time since they are already being nice in helping me then I try. I hope you understand my intent.

imann08 · #21 03-09-2007, 06:02 PM

Gary, I think we may be arguing different points with the importance of midtones. If I get this correctly, I think you are talking about it in terms of it's darkness as opposed to the highlight's correctness. In that case, I would think you are correct. I was talking in terms of color. If you have some highlight that you can be sure is not only neutral but also the lightest part of the image, then it is more important that the RGB are equal than it is for the midtones because once it is printed out, we will notice the slight differences more in the highlights. So when talking in terms of color, what the camera picks up really is irrelevant. In the case of how high or low those RGB values are, which is where I think you are coming from, then it would be correct that midtone are more reliable since a camera focuses on those at the expense of the highlights. That's why highlights are often blown out in images from digital cameras. Of course, you know that, but I am just typing what I'm thinking.

My knowledge of the inners of a camera as you explained are somewhere between slim and none. What you said is interesting and I'll have to think about it for a while. I hope you agree with what I said in the previous paragraph.

imann08 · #22 03-09-2007, 06:06 PM

albatross, thanks a lot. I'll take a look at that.

sunfly, that is an impressive correction. The only thing I disagree with is the removal of the background and how it leaves your image looking cut and pasted. That's not the important part of it though. At least as far as I can see, everything in the two people looks quite good. I'm impressed with the way the guy's hair came out and the baby's eye color. Very nice. Bravo!

Gary Richardson · #23 03-10-2007, 01:39 AM

Hi imann08,

Get where you're coming from now.

Yes, any colour imbalance will be more noticeable to the viewer in the high brightness areas, and so they must be correct.

However the best way to get colour balance for the whole picture in my opinion (and it is just an opinion) is still to get the mid-greys right first (for all the reasons previously stated), then do any final minor tweaks by balancing the highlights and shadows.

It may be that the final highlight adjustment "pulls" the mid-greys a bit, and that could be acceptable as long as the pic looks right, but adjusting the mid-greys first usually gets you into the ball park area faster. I've had some peculiar results when trying to balance things based solely on the highlights.

Usually I've had to adjust and readjust my curves before being satisfied that colour is balanced across all the lighting range, and there's almost always some compromise.

However based on what you say, I think I might experiment a bit more with letting the highlight settings predominate a little more than perhaps I have previously done.

imann08 · #24 03-10-2007, 02:10 AM

Well, I am glad we were able to clear that up. We were just on the wrong page. I can definitely see what you are talking about and I sometimes go straight for the midtones in images. It isn't a black and white thing and really depends on the image. If it is a high-key image then going for the highlights first is probably recommended as they have to keep detail. Of course, to back up what you were saying, most images will lie in the midtone areas. Generally speaking, I pay special focus on the highlights, set midtone points in the curve based on the area of focus, and the then adjust and finetune from there. That doesn't mean that I don't start by attacking midtones sometimes though. It has a lot to do with whether or not the detail is significant in the highlights areas and the fact that I am very conscious of making sure that the color of the highlights are perfect first and foremost.

I enjoyed this back and forth with you. I always look at it as a learning experience and know that you are quite experienced based on the responses I have seen you make on this board over time.

Gary Richardson · #25 03-10-2007, 07:43 AM

I too have enjoyed our chat and profited from the experience, it's by talking things through like this that we all learn, and after all that is what this site is all about.