Lon sent me a 100% crop of an oak tree atop a hill that was converted to B&W in PSCC-2017 using a Black& White adjustment layer. He wanted to darken cyans and blues in the sky to increase contrast between the clouds and open sky.
As you can see in the image, there is a jaggy white line near the intersection of the hillside and sky (red arrow).
A much larger version is Here
I decided to see if I could duplicate the issue, so here is what I did in a similar image that I made a few minutes later...
Create a Black & White adjustment layer above the background layer.
Set Blue and Cyan to -198 (as dark as possible).
The white line appears.
No other adjustments were made to that layer.
If I turn off the B&W layer, the white line disappears.
The image posted here, and in the linked image were sharpened using TK's Web Sharpen.
If any of you have an idea why this happens and/or have a solution, Lon and I would be most interested in knowing.
Right at the horizon line there is blue sky behind the grasses. Some of the blue is coming through the grasses, but it is not completely "blue" or cyan so the effect of darkening the sky is not being fully applied right at that edge.
Without seeing the original color image, that is a guess.
Curious why you also said set Blue/Cyan to -198 (as dark as possible). The sliders should go to -200.
Curious why you also said set Blue/Cyan to -198 (as dark as possible). The sliders should go to -200.
Oops. My bad. It should have said, "or, as dark as possible".
One thing I will check is to see at what value of blue and cyan the white line first appears.
Here's a shot of the oak--different composition, but the conditions are nearly identical.
--P
Preston, not sure I'm seeing or understanding the issue exactly, but there is a roof of a barn and a picket fence poking up in that general area of the horizon. It's very tiny, I wonder if it's the pickets in that fence that you're having an issue with?.......
Thanks for posting Preston. Yeah, this is a weird one I have not seen before. And before I go too much further, when I pushed the b&w adj layer to it's extreme, both cyan and blue about -157, it's probably going to cause some issues.... trouble is, I don't know what those issues are and more importantly, WHY it's happening.
On the lighter side, when a guy goes to the doctor complaining about a pain in his leg and tells the doctor, "My leg hurts when I move it like this...." The doc says, "Well, then don't move it like that." Kinda the same thing, if an anomaly shows up because you push a slider too far.... well, don't push the slider that far...!
STILL, it would be nice to understand what may be happening technically at the pixel level. Keith, you are correct, it's blue/cyan against green/yellow; with grasses and stuff extending up in to the blue sky area. Perhaps not all the blue and cyan pixels are getting selected in the adjustment layer.
Anyway, two more images. A closer zoom of the end result in b&w, and a similar zoom just from the opened RAW with few adjustments.
Paul, there is no barn or fence in the image. The white line in question is just below the hillside/sky interface, and the red arrow in my OP points to what is bugging Lon and I.
--P
........OK, I see the issue now. I thought it was right along the horizon and looked like the very tip of a roof line and tops of fence studs along that point. Closer in it is merely just sticks in the area I was referring too.
OK. I believe the issue is the Cyan/Blue controls in the B&W adjustment layer are just not fine tuned enough to solve the issue. It is doing what I suspected in my first post. Here's a quick version of the zoomed in piece that you posted in color. For this I used Select & Mask to get the sky, Used a Hue/Sat layer and set the saturation all the way down on the sky. Then darkened the sky using the same mask. Then inverted the mask and to deal with making the foreground B&W. I know this isn't a perfect selection, but the white halo is gone.
Thanks so much for chiming in. This little exercise is going to reveal how inept I am at masking and some of the finer details of PS. In fact, I had not used the "Select and Mask" feature yet and had to go watch some videos to learn how... :-(
I can get the selection, and even invert to create 2 b&w layers. I still get the line, but haven't been able to work in the desaturation. If I desaturate using the same selection mask, well, it applies to the whole sky, which defeats what I want to do with the sky. Sure, I can darken, but all contrast is lost. Yes, the line goes away, but still not sure how to just remove the line.
Still playing around. Kinda on a mission now. In Preston's famous words... "Research continues.."
Lon
Posted by John Williams on March 14, 2017 at 9:24 PM
I've never used a black and white adjustment layer, so this is an interesting thread.
I screen captured the color version, and played with darkening the cyan (and blue) as described. Pixel-peeking, I think the issue is Photoshop's definition of "cyan". Photoshop does not darken the pixels in the white line because they are not cyan enough.
In rgb, as opposed to lab, color is inseparable from luminosity. This means that the brighter a color is, the closer it approaches white. The problem here is that the color and luminosity both change as you move from the sky to the grass; the color from cyan to green (and yellow) and the luminosity from bright to dark. In that border region where the line develops, the luminosity is still quite bright as the color transitions from cyan to green.In other words the cyan color is lost as the green color is gained, and in that narrow region where the luminosity is still high Photoshop is not defining the color as cyan. Thus, when Photoshop darkens the cyan pixels, it does not recognize these "no longer cyan but still quite bright" pixels and leaves them as is.
What is fascinating to me is that if I darken all the sliders, the white line remains! This means that the transition is close enough to white that Photoshop does the same thing and leaves that same area bright even when green and yellow are darkened as well.
To sum, my simple brain thinks Photoshop is ignoring the white line because it is too neutral a color to be defined by whatever definintion Photoshop is using to apply the sliders.
Posted by Lon Overacker on March 15, 2017 at 2:18 PM
John,
Thank you very much for putting down your thoughts and explanations on this. What you're saying makes sense (well, at least to the degree I can understand it..)
thanks for chiming in.
Posted by John Williams on March 15, 2017 at 10:03 PM
In the spirit of a picture is a thousand words, here is an image with four colors of blue. (I realize they may not look "blue" as we normally call it, but that's a conversation for a different day .)
Moving from left to right, the rgb values are 0/0/255, 100/100/255, 200/200/255, and 254/254/255. Note that all four have the same blue value, only the red and green change from dark to light.
I'll post the BW conversion using Photoshop's BW layer below.
Posted by John Williams on March 15, 2017 at 10:09 PM
Here's the black and white. I converted this by adding a layer with all sliders at 0, except the blue which I set to -200. Note that the rgb values from left to right are now 0/0/0, 0/0/0, 90/90/90, and 225/225/225. (In each instance all values are equal, which is as should be for an image without color.) Note that Photoshop darkens the first two to black, the third one to a shade of gray a little under 50%, and the brightest blue a light shade of gray. This is despite the fact that each color block is 255 in the blue channel. Photoshop is essentially saying "I'll darken blue unless it is very bright, in which case I'll leave it bright." This works great for sky vs. clouds, but doesn't work well for that narrow transition from blue to green in the posted images.
How would I get around the issue? I'd steal one (or more) of the channels to base the image on.
Just for grins, I opened a Nikon RAW image in the latest version of ACR. I wondered if the same thing would happen, so I clicked on HSL/Greyscale and then adjusted the Blue slider to negative (darken blues) values. The same white line appears at -23 on the selected image, and gets progressively brighter as the negative value increases. (From Grins...to Groans)
I do not know if this is fixable by Adobe, but it would be nice if they could.
--P
Discussion
15 repliesAs you can see in the image, there is a jaggy white line near the intersection of the hillside and sky (red arrow).
A much larger version is Here
I decided to see if I could duplicate the issue, so here is what I did in a similar image that I made a few minutes later...
- The white line appears.
- No other adjustments were made to that layer.
- If I turn off the B&W layer, the white line disappears.
- The image posted here, and in the linked image were sharpened using TK's Web Sharpen.
If any of you have an idea why this happens and/or have a solution, Lon and I would be most interested in knowing.Thanks,
--P
Right at the horizon line there is blue sky behind the grasses. Some of the blue is coming through the grasses, but it is not completely "blue" or cyan so the effect of darkening the sky is not being fully applied right at that edge.
Without seeing the original color image, that is a guess.
Curious why you also said set Blue/Cyan to -198 (as dark as possible). The sliders should go to -200.
One thing I will check is to see at what value of blue and cyan the white line first appears.
Here's a shot of the oak--different composition, but the conditions are nearly identical.
--P
On the lighter side, when a guy goes to the doctor complaining about a pain in his leg and tells the doctor, "My leg hurts when I move it like this...." The doc says, "Well, then don't move it like that." Kinda the same thing, if an anomaly shows up because you push a slider too far.... well, don't push the slider that far...!
STILL, it would be nice to understand what may be happening technically at the pixel level. Keith, you are correct, it's blue/cyan against green/yellow; with grasses and stuff extending up in to the blue sky area. Perhaps not all the blue and cyan pixels are getting selected in the adjustment layer.
Anyway, two more images. A closer zoom of the end result in b&w, and a similar zoom just from the opened RAW with few adjustments.
--P
Thanks so much for chiming in. This little exercise is going to reveal how inept I am at masking and some of the finer details of PS. In fact, I had not used the "Select and Mask" feature yet and had to go watch some videos to learn how... :-(
I can get the selection, and even invert to create 2 b&w layers. I still get the line, but haven't been able to work in the desaturation. If I desaturate using the same selection mask, well, it applies to the whole sky, which defeats what I want to do with the sky. Sure, I can darken, but all contrast is lost. Yes, the line goes away, but still not sure how to just remove the line.
Still playing around. Kinda on a mission now. In Preston's famous words... "Research continues.."
Lon
I screen captured the color version, and played with darkening the cyan (and blue) as described. Pixel-peeking, I think the issue is Photoshop's definition of "cyan". Photoshop does not darken the pixels in the white line because they are not cyan enough.
In rgb, as opposed to lab, color is inseparable from luminosity. This means that the brighter a color is, the closer it approaches white. The problem here is that the color and luminosity both change as you move from the sky to the grass; the color from cyan to green (and yellow) and the luminosity from bright to dark. In that border region where the line develops, the luminosity is still quite bright as the color transitions from cyan to green.In other words the cyan color is lost as the green color is gained, and in that narrow region where the luminosity is still high Photoshop is not defining the color as cyan. Thus, when Photoshop darkens the cyan pixels, it does not recognize these "no longer cyan but still quite bright" pixels and leaves them as is.
What is fascinating to me is that if I darken all the sliders, the white line remains! This means that the transition is close enough to white that Photoshop does the same thing and leaves that same area bright even when green and yellow are darkened as well.
To sum, my simple brain thinks Photoshop is ignoring the white line because it is too neutral a color to be defined by whatever definintion Photoshop is using to apply the sliders.
Thank you very much for putting down your thoughts and explanations on this. What you're saying makes sense (well, at least to the degree I can understand it..)
thanks for chiming in.
Moving from left to right, the rgb values are 0/0/255, 100/100/255, 200/200/255, and 254/254/255. Note that all four have the same blue value, only the red and green change from dark to light.
I'll post the BW conversion using Photoshop's BW layer below.
How would I get around the issue? I'd steal one (or more) of the channels to base the image on.
I do not know if this is fixable by Adobe, but it would be nice if they could.
--P