Htydjtk

The skin colour of my subject is widely inconsistent and I am not sure what the reason for it is. To me it looks like the side of the face suffers from a colour shadow thrown by the bright hair. At the same time, the skin colour closer towards the centre of the face seems like an exaggeration of minor differences in colour of the subject's skin. Those reddish/pink spots aren't really visible to my eye.

What is the best approach to avoid such issues as I am taking a photo so I don't have to try and fix it in Photoshop afterwards? Is it something with my settings?

Picture is taken in JPG in manual mode with automatic white balance and spot-metering. No filters or special camera modes.

You are likely not getting the colors you want because of the color profile on your camera or raw processing software does not match your preferences. Camera and lens selection may also play a role.

Lens Selection

Lenses may transmit different frequencies differently. For instance, some produce warmer colors, while others are cooler. Some lenses also have defects, such as "glow", when shot wide open. Some people find that a soft-focus look is pleasant in portraits.

Camera Settings

If you are disciplined, you should set custom white balance. However, if you shoot in conditions with varied lighting, forgetting to change white balance can result in dozens of subsequent shots being ruined. Some cameras allow color-shift adjustments to auto white balance. Since AWB on my camera tends to produce images with more magenta than I'd like, I adjust AWB to increase the complementary color, green.

You should also select the color profile on your camera that best matches your preferences. Usually one or two profiles will subdue magenta-red colors. Available options vary by camera maker:

• 
  FujiFilm: Provia, Velvia, Astia, Classic Chrome, ProNeg-High, ProNeg-Low.


• 
  Canon: Standard, Portrait, Landscape, Neutral, Faithful.


• 
  Nikon: Standard, Neutral, Vivid, Portrait, Landscape, Flat.


• 
  Olympus (Picture Modes): i-Enhance, Vivid, Natural, Muted, Portrait.



• Sony (Creative Style): Standard, Vivid, Neutral, Clear, Deep, Light, Portrait, Landscape, Sunset, Night Scene, Autumn leaves.




• _{(Feel free to suggest edits with camera makers and profile names...)}

Cameras also typically allow adjustments to saturation, contrast, highlight, shadows, noise reduction, and sharpness.

Post Processing

The objectionable skin tones are most likely magenta, not red. Even if the colors are accurate, you may still prefer less magenta.

• You can reduce the magenta by adjusting curves to increase the complementary color, green, in relevant areas. Use color-layer blending to avoid changing the overall luminosity of the image. Use layer masks to isolate changes.




• Sometimes desaturating slightly is enough to fix skin tones. Use a layer mask to isolate changes.




• Some editors include skin-tone specific tools. If these are available to you, experiment with them until you find settings you like.

In your sample image, the "whites" of the girl's eyes are pinkish. Here is the image after adjusting the gamma of the green (1.15) and red (0.85) channels to neutralize the color cast somewhat. Her cheeks are still pinkish, but it should appear more natural and perhaps more to your liking.

It's certainly a very warm colour balance. Is this more accurate? It's a simple 'Auto Color' process, available in most photo editors.

But, although it's refreshing to see a young girl not plastered in makeup, it IS rather cruelly accurate.

What shall we do about the shadow under the cheekbone? Maybe do some subtle work with the airbrush. Or just re-shoot, but reflect some light into that area?

What kind of light was illuminating your subject?

It all starts with the light.

Always.

I've found that most of the time uneven skin tones are a result of:

• Too many different types of lighting mixed from different angles, so that one part of skin is influenced more by, say, the tungsten lamp to the right while another part of the skin is more influenced by the fluorescent light high overhead to the left. Color casts from highly reflective objects, such as your subject's pink wig or even a rich green lawn on a sunny day, can shift the color of human skin.


• Poor overall lighting, even if all of it is the same type, with a low CRI (color rendering index). If there is no light present that is the same hue as parts of your subject's skin, there will be no light that color reflected from your subject. The colors that are present in the light will then be exaggerated.


• Flickering lighting, such as most "energy efficient" light sources like cheap fluorescent or LED lighting, combined with fairly fast exposure times so that as the opening between the shutter curtains transits the sensor, the intensity and color of the light is changing as the light cycles from the peak to the trough of the alternating current powering it. Flickering lights not only get brighter and dimmer with the cycle of the current powering them, they also go from bluer and fuller spectrum at peak to browner and lower spectrum at the trough.

What is the best approach to avoid such issues as I am taking a photo so I don't have to try and fix it in Photoshop afterwards? Is it something with my settings?

Control your light sources. Be sure you are using even, full spectrum lighting that does not flicker. Be sure that all of the light coming from different angles is the same, both in terms of color temperature along the amber ←→ blue axis and "tint" along the green ←→ magenta axis.

Control your camera's color. Be sure the camera's white balance setting matches your light. Again, this includes both color temperature adjustment as well as white balance correction along the "tint" axis.

Shoot raw anyway. As counterintuitive as this sounds, saving all of the raw data will allow you to make a few, simple global adjustments in post processing to dial your color in even more precisely than the coarser settings on your camera will allow. You won't be forced to do various localized adjustments to a JPEG image near as often by processing the raw data.