Neutralizing color cast in shadows on white uniforms

troutstreaming · March 8, 2008

I have been scratching my head for the right combination of masks/blend-if's/channel blends/curves etc. to neutralize colorcasts in the shadow areas on white uniforms and am hoping that someone out there has some suggestions.

When shooting from under the basket at a local college the light placement on the ceiling leads to a drop off in direct light on the edges of the court, so the players are lit from above (and behind) with the reflected light from the wall/stands/floor providing fill and it is usually carrying a color cast. As the home team is in white it is very evident and exists in a wide range of luminosity values. Preserving skin tones is also important so I need to be cognizant of not tweaking them in the process (examples below). My normal workflow comes from the Dan Margulis books and uses a once for color once for contrast approach with by the numbers RGB individual channel work to create a good tonal foundation, followed by LAB curves or apply image's to increase saturation and I frequently employ channels as masks and use blend-if's in different color spaces to restrict changes to specific areas so I am willing to try anything, but my imagination/vision just has not been great enough to tackle this color cast issue (and my pockets not deep enough to install stobes on the ceiling of the gym - no on camera flash would be allowed).

Thank you in advance for your suggestions,

Both of the following images have been post-processed using my normal workflow.

15 feet from the baseline everything is manageable -

But under the rim the combination of lighting is too much for my skill set - in this image I neutralized the direct light, but did not even attempt to netralize the shadows (more accurately I did not bother to save any of the attempts at neutralizing the shadow casts...) -

jfriend · March 8, 2008

troutstreaming wrote:

I have been scratching my head for the right combination of masks/blend-if's/channel blends/curves etc. to neutralize colorcasts in the shadow areas on white uniforms and am hoping that someone out there has some suggestions.

When shooting from under the basket at a local college the light placement on the ceiling leads to a drop off in direct light on the edges of the court, so the players are lit from above (and behind) with the reflected light from the wall/stands/floor providing fill and it is usually carrying a color cast. As the home team is in white it is very evident and exists in a wide range of luminosity values. Preserving skin tones is also important so I need to be cognizant of not tweaking them in the process (examples below). My normal workflow comes from the Dan Margulis books and uses a once for color once for contrast approach with by the numbers RGB individual channel work to create a good tonal foundation, followed by LAB curves or apply image's to increase saturation and I frequently employ channels as masks and use blend-if's in different color spaces to restrict changes to specific areas so I am willing to try anything, but my imagination/vision just has not been great enough to tackle this color cast issue (and my pockets not deep enough to install stobes on the ceiling of the gym - no on camera flash would be allowed).

Thank you in advance for your suggestions,

Both of the following images have been post-processed using my normal workflow.

15 feet from the baseline everything is manageable -

But under the rim the combination of lighting is too much for my skill set - in this image I neutralized the direct light, but did not even attempt to netralize the shadows (more accurately I did not bother to save any of the attempts at neutralizing the shadow casts...) -

Check out this thread which is about the color of snow, but I think relevant to your question. White reflects the light that hits it. When light hits a white surface directly, our eye expects it to "look white" even if the light isn't quite white. When a white surface is not receiving direct light, it is reflecting the reflected light that it gets which will be colored by whatever it's reflected off of. In a basketball game, this will most likely be influenced by the color of the floor. If you had a shot that was 100% shadows, then you would probably want to neutralize the shadows because that's generally how the eye/brain will process it in person. But, if you have a mix of shadows, the eye/brain expects the directly lit whites to be white and it's OK for the shadows to have a color cast from the environment (probably a warm cast in the case of a basketball floor).

The 2nd shot in your posting looks good to me. The directly lit white are very nearly neutral. The shadow whites are very warm (reflected light from the floor) and the skin tone numbers look believable. If the uniforms are actually white, this looks pretty good to me.

Because the shadows are easily separable from the highlights in a curve, it is possible to construct a curve that will move the shadows more towards neutral (less red, more blue) while not affecting the rest of the image, though I don't think it's necessarily needed here because the eye/brain is often OK with highlights and shadows have slightly different colors since that happens in nature all the time.

troutstreaming · March 8, 2008

jfriend wrote:

Check out this thread which is about the color of snow, but I think relevant to your question. White reflects the light that hits it. When light hits a white surface directly, our eye expects it to "look white" even if the light isn't quite white. When a white surface is not receiving direct light, it is reflecting the reflected light that it gets which will be colored by whatever it's reflected off of. In a basketball game, this will most likely be influenced by the color of the floor. If you had a shot that was 100% shadows, then you would probably want to neutralize the shadows because that's generally how the eye/brain will process it in person. But, if you have a mix of shadows, the eye/brain expects the directly lit whites to be white and it's OK for the shadows to have a color cast from the environment (probably a warm cast in the case of a basketball floor).

The 2nd shot in your posting looks good to me. The directly lit white are very nearly neutral. The shadow whites are very warm (reflected light from the floor) and the skin tone numbers look believable. If the uniforms are actually white, this looks pretty good to me.

Because the shadows are easily separable from the highlights in a curve, it is possible to construct a curve that will move the shadows more towards neutral (less red, more blue) while not affecting the rest of the image, though I don't think it's necessarily needed here because the eye/brain is often OK with highlights and shadows have slightly different colors since that happens in nature all the time.

Thanks for taking the time to look and comment and for the reassurance that the 2nd image is acceptable. My personal preference though is to try and tone down the cast more which is what I am struggling with. While in the setting of the gym the physics of the situation certainly creates a cast and our perception of the scene may account for the casts, I think that the image would be improved if it was less noticable and that if you polled those present watching the game few if any would 'remember' the presence of the cast under the hoop in the indirectly lit areas.

Unfortunately if you look at the luminosity values of the image the directly lit areas range in L value from 98 down to the 80's for 'whites' with the 'blacks' in the teens, while the indirectly lit regions of the image range from 'whites' in the 60's to blacks at 0-2. The maroon in the uniform ranges from the mid-teens to low 20's when directly lit to 2-15 when indirectly lit. The skin tones range from 70-90 in the directly lit areas to 30-45 when indirectly lit. Please note that I am intentionaly avoiding the use of 'shadow' and 'highlight' here when refering to the luminance of the image. A straight luminosity mask approach over a curve that would neutralize the indirect cast would need to reach from 0 - 60 while allowing the strongest changes in the range of the 'white' neutral values in the 60's (placing it very close to the direclty lit skin tones) and this would also affect the directly lit blacks and maroons. I have had no luck constructing such masks, nor with adding combinations of blend-if's based on color channel information. Maybe desaturating all of the the 'shadow' part of the luminosity mask (affecting both the directly and indirectly lit maroons and blacks) while trying to restrict the indirect light correction curve to the 30-65 L range so that it most affects the indirectly lit skin tones and whites would help but I am tired of playing with this image currently so I thought that I would toss this question out there for others opinions/suggestions.

Thanks again for looking and commenting - I will try and give the 'snow' thread a read and will reconsider my desire to further neutralize the indirectly lit whites.

jfriend · March 8, 2008

troutstreaming wrote:

Thanks for taking the time to look and comment and for the reassurance that the 2nd image is acceptable. My personal preference though is to try and tone down the cast more which is what I am struggling with. While in the setting of the gym the physics of the situation certainly creates a cast and our perception of the scene may account for the casts, I think that the image would be improved if it was less noticable and that if you polled those present watching the game few if any would 'remember' the presence of the cast under the hoop in the indirectly lit areas.

Unfortunately if you look at the luminosity values of the image the directly lit areas range in L value from 98 down to the 80's for 'whites' with the 'blacks' in the teens, while the indirectly lit regions of the image range from 'whites' in the 60's to blacks at 0-2. The maroon in the uniform ranges from the mid-teens to low 20's when directly lit to 2-15 when indirectly lit. The skin tones range from 70-90 in the directly lit areas to 30-45 when indirectly lit. Please note that I am intentionaly avoiding the use of 'shadow' and 'highlight' here when refering to the luminance of the image. A straight luminosity mask approach over a curve that would neutralize the indirect cast would need to reach from 0 - 60 while allowing the strongest changes in the range of the 'white' neutral values in the 60's (placing it very close to the direclty lit skin tones) and this would also affect the directly lit blacks and maroons. I have had no luck constructing such masks, nor with adding combinations of blend-if's based on color channel information. Maybe desaturating all of the the 'shadow' part of the luminosity mask (affecting both the directly and indirectly lit maroons and blacks) while trying to restrict the indirect light correction curve to the 30-65 L range so that it most affects the indirectly lit skin tones and whites would help but I am tired of playing with this image currently so I thought that I would toss this question out there for others opinions/suggestions.

Thanks again for looking and commenting - I will try and give the 'snow' thread a read and will reconsider my desire to further neutralize the indirectly lit whites.

You got me curious on how to do this one. It is difficult to isolate just the white of the uniform in either tone or color. The problem is that there are so many different colors in play all with overlapping tones that I couldn't seem to isolate the cast color in the uniform, even trying with LAB values. There's probably a more elegant way to do it, but here's what I did:

I create a hue/saturation adjustment layer.
I desaturate almost all the way to completely remove the cast.
I use a blend-if layer to bring the color back to the darkest tones. At this point, I'm just trying to get the color back in the numbers on the uniform, while keeping the color out of the whites in the uniform. I'm ignoring the rest of the image.
Then, I hand painted a rough mask to block the desaturation from the rest of the image (everything other than the whites of the uniform). With a soft brush, this could be pretty fast and I didn't need to do any detail work.
Lastly, I had to deal with the hand that protrudes across the white uniform. I'm sure there are more elegant ways to do this, but I just used the polygon selection tool and then a slight feather. Since it's just desaturation and the color in the whites isn't that big a deal as long as you feather and cover the whole arm, this also doesn't have to be really precise.

I ended up with this:

Compared to the one with the cast:

I would probably vote for a slight blend of the two, but at this point, it's personal taste. Perhaps worth doing for a real money shot. Certainly not something you could afford to do on a whole shoot.

troutstreaming · March 8, 2008

jfriend wrote:

You got me curious on how to do this one. It is difficult to isolate just the white of the uniform in either tone or color. The problem is that there are so many different colors in play all with overlapping tones that I couldn't seem to isolate the cast color in the uniform, even trying with LAB values. There's probably a more elegant way to do it, but here's what I did:
I create a hue/saturation adjustment layer.

I desaturate almost all the way to completely remove the cast.

I use a blend-if layer to bring the color back to the darkest tones. At this point, I'm just trying to get the color back in the numbers on the uniform, while keeping the color out of the whites in the uniform. I'm ignoring the rest of the image.

Then, I hand painted a rough mask to block the desaturation from the rest of the image (everything other than the whites of the uniform). With a soft brush, this could be pretty fast and I didn't need to do any detail work.

Lastly, I had to deal with the hand that protrudes across the white uniform. I'm sure there are more elegant ways to do this, but I just used the polygon selection tool and then a slight feather. Since it's just desaturation and the color in the whites isn't that big a deal as long as you feather and cover the whole arm, this also doesn't have to be really precise.

I ended up with this:

Compared to the one with the cast:

I would probably vote for a slight blend of the two, but at this point, it's personal taste. Perhaps worth doing for a real money shot. Certainly not something you could afford to do on a whole shoot.

Thank you for taking the time and making the effort! I agree that the final result should probably be a blend of the two, but once you have both to play with its just a matter of moving sliders.

So the magic trick then is to try and figure out some sort of calculated layer masking to restrict the desaturated layer to objects that are more 'tainted' white then tainted 'flesh tone' (or wall or ceiling...) to use instead of the handpainted layer mask and use that to mask your desaturation/blend-if. Where's the artifiical intelligence when you need it...

Thanks again!

Neutralizing color cast in shadows on white uniforms

Comments