The Color of Snow

rutt · January 26, 2005

You might have thought we were done with this topic, but with so much snow on the ground here and more fallling today, the topic remains topical and I've learned something more about it from a member of the Color Theory mailing list, Steven Barton.

Here is a dog in the snow. This is the out-of-camera jpeg, but the sky was overcast and the snow in the shot is pretty close to neutral:

It's OK, but there is no detail in the snow. The simplest thing is just to neutralize the snow and blow it out. But let's try something different:

Better? What I did was bring up the detail in the snow (the opposite of blowing it out) by steepening the L curve in the area of the snow:

In this picture, the highlight is in the snow and the highlight is too dark. So I was able to move the highlight point to the right, in effect stealing some contrast for the area of snow that actually has detail. This part of the curve I made very steep and it brought up the detail in the snow nicely. Then I flattened the curve through the midtones because in this shot there is really nothing there and I wanted to save contrast for the dog. I used this detail by steepening the curve gain through the shadows. I finished it with the FM CSPro sharpening tool. Better snow, eh? And better dog.

Here is another example. Here is the original out-of-the-camera jpeg:

Too blue and too dark. Here is the version I made on Sunday by neutralizing and blowing out the snow (somewhat):

Better, but I think there is more to get from this image. Here is a version I made this morning:

Here is the L curve that I used:

In this case I didn't have the luxury to completely blow out the highlights because the real highlight of this shot is the sky and it has to be lighter than the snow and also will look strange against the trees if it's too blown. So instead, I flattened the curve through the sky section and steepened it through the snow section of the highlights. I flattened again and then steepened in the midtownes where the kid's face is to bring up the detail there. I also did a little simple steepening of the A and B curves to increase the saturation (as long as I was doing LAB curves.) BTW, in this and the following cases I started out by using the new trick I learned from you all, custom white balance in the adobe raw converter.

Here are a couple of other examples of basically the same thing.

Out-of-camera jpeg:

Sunday's edit:

Today's version:

The L curve for today's version:

Out-of-camera jpeg:

Sunday's version:

Today's version:

Today's L curve:

Hope I haven't beaten a dead horse, but this has been very educational for me. (And I hope others.)

DoctorIt · January 26, 2005

It is still beyond me how you pick those intermediate points in your curves.

Your results look really good though, nice!

rutt · January 26, 2005

DoctorIt wrote:

It is still beyond me how you pick those intermediate points in your curves.

It isn't random and it isn't even experience; there is a trick. With the curve open, push the mouse (left mouse) button down while you move the mousee cursor over the image. A point appears on the curve to show the exact position on the curve of the pixel you point at. So I was easily able to find the lightest and darkest points in the curve by pointing at the lightest and darkest points in the snow. This is a lot easier than reading out the color sampler values ahead of time and entering them into the curves, which also works, btw.

There is another related trick. With the curves dialog open, Command-Click (mac, I assume Alt-Click on windows) and a new point will be added to the curve at the point with the current input of the pixel you are pointing at. (What a bad sentence! But try it, open a curves dialog and Command-Click or Alt-Click on some point in the image.)

Writing good L curves is easy (compared to writing good curves in separate CMYK channels.) Figure out where in the image you want detail (contrast), use the aforementioned trick to find that relevant part of the curve, and make it steep there. Then figure out how to pay for this steepness buy flattening the curve in the parts of the image that don't need the detail as much.

cletus · January 26, 2005

rutt wrote:

There is another related trick. With the curves dialog open, Command-Click (mac, I assume Alt-Click on windows) and a new point will be added to the curve at the point with the current input of the pixel you are pointing at. (What a bad sentence! But try it, open a curves dialog and Command-Click or Alt-Click on some point in the image.)

rutt,

I think it's Ctrl + Click on the PC. For just about all of the Photoshop keyboard shortcuts I've seen Command on the Mac corresponds to Ctrl on the PC and Option on the Mac corresponds to Alt on the PC.

DoctorIt · January 26, 2005

hmm, I'll give that a try (i'm on a mac, but thanks cletus!). I have been using your steepening tricks from a while ago in the A and B channels, but I've stayed away from the L... cause you told me to back then when we were talking about making colors pop.

Thanks!

DavidTO · January 26, 2005

DoctorIt wrote:

back then when we were talking about making colors pop.

Where's that thread?

rutt · January 26, 2005

DavidTO wrote:

Where's that thread?

I've written a couple of things about LAB curves. I think the best thing I posted was http://www.dgrin.com/showthread.php?t=2042 Maybe Erik remembers a different post.

Really everything I know about LAB curves (and almost everything I know about color correction) I learned from Dan Margulis. If you can get through The Magic of LAB, you'll know at least as much as I do.

pathfinder · January 26, 2005

rutt wrote:

You might have thought we were done with this topic, but with so much snow on the ground here and more fallling today, the topic remains topical and I've learned something more about it from a member of the Color Theory mailing list, Steven Barton.

Here is a dog in the snow. This is the out-of-camera jpeg, but the sky was overcast and the snow in the shot is pretty close to neutral:

It's OK, but there is no detail in the snow. The simplest thing is just to neutralize the snow and blow it out. But let's try something different:

Better? What I did was bring up the detail in the snow (the opposite of blowing it out) by steepening the L curve in the area of the snow:

In this picture, the highlight is in the snow and the highlight is too dark. So I was able to move the highlight point to the right, in effect stealing some contrast for the area of snow that actually has detail. This part of the curve I made very steep and it brought up the detail in the snow nicely. Then I flattened the curve through the midtones because in this shot there is really nothing there and I wanted to save contrast for the dog. I used this detail by steepening the curve gain through the shadows. I finished it with the FM CSPro sharpening tool. Better snow, eh? And better dog.

Here is another example. Here is the original out-of-the-camera jpeg:

Too blue and too dark. Here is the version I made on Sunday by neutralizing and blowing out the snow (somewhat):

Better, but I think there is more to get from this image. Here is a version I made this morning:

Here is the L curve that I used:

In this case I didn't have the luxury to completely blow out the highlights because the real highlight of this shot is the sky and it has to be lighter than the snow and also will look strange against the trees if it's too blown. So instead, I flattened the curve through the sky section and steepened it through the snow section of the highlights. I flattened again and then steepened in the midtownes where the kid's face is to bring up the detail there. I also did a little simple steepening of the A and B curves to increase the saturation (as long as I was doing LAB curves.) BTW, in this and the following cases I started out by using the new trick I learned from you all, custom white balance in the adobe raw converter.

Here are a couple of other examples of basically the same thing.

Out-of-camera jpeg:

Sunday's edit:

Today's version:

The L curve for today's version:

Out-of-camera jpeg:

Sunday's version:

Today's version:

Today's L curve:

Hope I haven't beaten a dead horse, but this has been very educational for me. (And I hope others.)

John, this is a very educational discussion. The snow has lost its blue-cyan cast in each finished image and the skin tones are dramatically better - warmer and more neutral and sun lit looking. Is the improvement in the images a result of the LAB curves, or due to the dramatically better color balance due to the use of the white balance eyedropper in ARC? The improvement in color balance is oobvious. The increased tonality in the snow is more subtle on my monitor.

I read the pixel data in my snow scenes and if they are not a very light neutral gray, I tend to smell a ringer. There should be no predominance of Red, Green, or Blue in snow unless the sun is not shining - before sunrise, after sunset, overcast, etc or a color is reflected onto the snow from a colored wall or something.

DavidTO · January 26, 2005

rutt wrote:

I've written a couple of things about LAB curves. I think the best thing I posted was http://www.dgrin.com/showthread.php?t=2042 Maybe Erik remembers a different post.

Really everything I know about LAB curves (and almost everything I know about color correction) I learned from Dan Margulis. If you can get through The Magic of LAB, you'll know at least as much as I do.

I've got the book, but I guess it doesn't do me any good if I don't read it.

Thanks.

rutt · January 26, 2005

:uhoh

DavidTO wrote:

I've got the book, but I guess it doesn't do me any good if I don't read it.

Thanks.

The Magic of LAB article is a better intro to LAB than anything in the book.

rutt · January 26, 2005

pathfinder wrote:

John, this is a very educational discussion. The snow has lost its blue-cyan cast in each finished image and the skin tones are dramatically better - warmer and more neutral and sun lit looking. Is the improvement in the images a result of the LAB curves, or due to the dramatically better color balance due to the use of the white balance eyedropper in ARC? The improvement in color balance is oobvious. The increased tonality in the snow is more subtle on my monitor.

Yeah, the improved snow contrast is easier to see in PS and I suspect it will print a lot better. The ones from today use both ARC white balance and also symetrical LAB curve steepening. I think that starting with ARC white balanced originals made it very easy to write better A+B curves. Some of the sunlit look comes from the L curves which lighten the images and get the contrast where it counts (snow and face or dog.)

To my eye, I think the ones from today are just a little too warm on the snow. It really is a little blue. So I suppose there is still a little fun left to be had here.

pathfinder · January 26, 2005

rutt wrote:

Yeah, the improved snow contrast is easier to see in PS and I suspect it will print a lot better. The ones from today use both ARC white balance and also symetrical LAB curve steepening. I think that starting with ARC white balanced originals made it very easy to write better A+B curves. Some of the sunlit look comes from the L curves which lighten the images and get the contrast where it counts (snow and face or dog.)

To my eye, I think the ones from today are just a little too warm on the snow. It really is a little blue. So I suppose there is still a little fun left to be had here.

I'll have to try some of these techniques when I have new snow shots to process.

I am reading Ben Wilmore's Photoshop CS Studio Techniques and he talks at length about adjusting curves - and you don't really have to go to LAB, you can adjust curves and the Fade with Luminosity as alternative. I think I am beginning to finally get a handle on curve's. Wilmore's book is excellent and easy to read. I suspect you would like it.

rutt · February 8, 2005

rutt wrote:

To my eye, I think the ones from today are just a little too warm on the snow. It really is a little blue. So I suppose there is still a little fun left to be had here.

I had this fun. Now I move the B curve just a touch (1 or 2 points) toward blue by moving the blue endpoint inward. So this is assymetrical LAB B steepening. It gives the snow just a tad of a blue cast. Sometimes, I keep the yellow side of the curve constant so as to keep the faces &etc warm. But I think the blue snow and sky create a slight overall blue cast. The images treated this way look very natural to my eye. For example:

and:

wxwax · February 8, 2005

Great snow and great info, Rutt. I read a thing yesterday that explained why snow is white and ice is blue.

Light waves don't get very far into snow before they start bouncing around like crazy. As a result, they don't get filtered out by wavelength. So white light is reflected back.

But ice is denser, so light waves don't bounce around. The wavelength which penetrates the furthest is blue. So that's why icebergs seem to have a blue hue.

I think I got that right.

Joolya · February 10, 2005

wxwax wrote:

Great snow and great info, Rutt. I read a thing yesterday that explained why snow is white and ice is blue.

As mentioned in the "What color is snow thread:" Yes, that is correct. So it makes sense that if you mess with the red saturation the picture will look more like the snow we percieve. When our flash is hitting the snow, the blue is saturating the most and that's why snow in pictures often turns out blue. I'm sure a lot has to do with reflection and what not but that is the more scientific description. Check this out: http://www.discovery.com/area/skinn...3/skinnyon.html

wxwax · February 12, 2005

One more link on how to expose for snow.

http://www.kenrockwell.com/tech/snow.htm

The Color of Snow

Comments