Fun With Highlight Testing/Recovery!

I might have it backwards. It's not something I tried too hard to think about, when I was testing it.

In LCD's, the pixels alternate on/off with neighboring pixels. I guess the faster shutter speeds will freeze it in one of those states (giving you that pattern). And shutter speeds below the monitor's refresh will make it less apparent since it's going faster relative to the shutter.

Str0ntium wrote in post #13369753
I might have it backwards. It's not something I tried too hard to think about, when I was testing it.

In LCD's, the pixels alternate on/off with neighboring pixels. I guess the faster shutter speeds will freeze it in one of those states (giving you that pattern). And shutter speeds below the monitor's refresh will make it less apparent since it's going faster relative to the shutter.

Hmm, well, the shot I took at 1/15 sec doesn't show any, so you may be right there!

The lines are just a moiré pattern created by the pixel grid in the LCD.

Hey again, gang!

Well, I did another "test project" today! This time, an outdoors (from my deck) scene of an overcast day but with some light sky, and a pretty dim foreground with some fall colors.

If you are interested and want to see the PBase "full description" with all my shots without first reading this little intro, here's the link:

http://www.pbase.com …11_overcast_day​_exposures

Well, then, for my first shot I decided to "Expose for the highlights" by setting my 5D to expose a brighter area of the sky to, I believe, +1 1/3EV:

Tony, 80% of the apparent clipping you see in LR (or on the camera review) is caused by white balancing. It is much more a factor than contrast, saturation or color space settings. To know, in LR, if your image is really clipped before WB is applied, temporarily set Temperature to 2050 and Tint to -150. At this setting the WB multiplier values are x1, in other words WB is effectively cancelled and you have a histogram that is gamma corrected but not white balanced.
Why is knowing this useful? Because if there is no capture clipping, only WB induced clipping, it is better to stay away from Recovery when editing. Recovery flattens the tone curve applied to the image at its top. This means a loss of contrast and detail in the highlights. Of course, if there is capture clipping, Recovery is needed because as well as flattening the tone curve it also interpolates data from the unclipped channel(s) to the clipped one(s). But if there is no real clipping, reducing Exposure is a better way of counteracting WB clipping because its action is linear. After the white point has been pulled down enough to get rid of the clipping, a combination of Brightness, Contrast and Fill will restore levels to where they need to be without moving the white point. (Even if Brightness or Contrast cause the clipping alarm to go red, the image isn't really clipped. The alarm is activated before 255.) But beware of the Adobe Standard profile; it also can cause false clipping.

Here's an interesting note: Eric Chan (LR/CR design team member) revealed on the Luminous Light forum some time ago that a minus Exposure setting triggers the application of some recovery even when Recovery is set to 0. But this is conditional on there being actual clipping. Without it Exposure remains linear.

tzalman wrote in post #13374475
Tony, 80% of the apparent clipping you see in LR (or on the camera review) is caused by white balancing. It is much more a factor than contrast, saturation or color space settings. To know, in LR, if your image is really clipped before WB is applied, temporarily set Temperature to 2050 and Tint to -150. At this setting the WB multiplier values are x1, in other words WB is effectively cancelled and you have a histogram that is gamma corrected but not white balanced.
Why is knowing this useful? Because if there is no capture clipping, only WB induced clipping, it is better to stay away from Recovery when editing. Recovery flattens the tone curve applied to the image at its top. This means a loss of contrast and detail in the highlights. Of course, if there is capture clipping, Recovery is needed because as well as flattening the tone curve it also interpolates data from the unclipped channel(s) to the clipped one(s). But if there is no real clipping, reducing Exposure is a better way of counteracting WB clipping because its action is linear. After the white point has been pulled down enough to get rid of the clipping, a combination of Brightness, Contrast and Fill will restore levels to where they need to be without moving the white point. (Even if Brightness or Contrast cause the clipping alarm to go red, the image isn't really clipped. The alarm is activated before 255.) But beware of the Adobe Standard profile; it also can cause false clipping.

Elie, that's interesting! I've never waded into the White Balance "stuff". But, I could take the one image I got yesterday that didn't clip in the camera but does in LR and play with the WB in LR and see what it does to the clipping, post about it here later!

tzalman wrote in post #13374482
Here's an interesting note: Eric Chan (LR/CR design team member) revealed on the Luminous Light forum some time ago that a minus Exposure setting triggers the application of some recovery even when Recovery is set to 0. But this is conditional on there being actual clipping. Without it Exposure remains linear.

This is also interesting -- does this mean that when there is clipping pulling back on Exposure emphacises hightlights?

Of course I've heard of problems with Recovery, but I'm not clear on what the difference is between Recovery and pulling back on Highlights in the Tonal Curve controls is...?

I pretty much just go by "look and feel"...!

I see that the image posted above was converted with Adobe Standard. Put it into Camera Standard or Camera Neutral and watch what happens to the histogram.

tzalman wrote in post #13377756
I see that the image posted above was converted with Adobe Standard. Put it into Camera Standard or Camera Neutral and watch what happens to the histogram.

I'll try that!

OK, I took the last "overexposed" (not really) image before adjusting it/taming it. The brighter areas of the sky still showed clipping in LR. So, there were just subtle changes but not enough to make the clipping go away.

Camera Faithful seemed to do the best job with the darker clouds.

And then, I tried Camera Standard and Camera Landscape, and they really perked up the colors! In fact, using Camera Landscape I actually had to tone down the reds a bit!

Here's the result:

When shooting with Uni-WB I first check the LR histogram to be sure there is no clipping (my default Basic settings are everything zeroed and Neutral profile) and then WB by clicking a neutral target (or one I want to be neutral, like clouds which aren't really white or grey, but pictorally I like them to be white) or by using a preset and then tweaking it. I don't like LR's Daylight preset (5500 K), too warm for my taste, so I have my own User Preset for WB at 4900/+3.
Then I hit Auto Tone to see where LR puts the white and black points. If I know the image will be exported in sRGB, I pull the white point down (minus Exposure) to gain some headroom because I am always afraid of the smaller gamut causing clipping. Finally, I readjust Fill, Brightness and Contrast.

When I am shooting an event where people may ask to see the camera LCD, I don't use Uni-WB. Instead I use a temperature based WB of 4300 K. At this temperature the red and blue channels are boosted to near parity with the green, but not much more, so there is less chance of WB clipping. The color may not be great, but it's less alarming than the green Uni-WB. In this case (if I remember) I put on an import preset that drops the WB to 2050/-150, the equivalent of Uni-WB, and then continue as above.

Well, gang, I'm not sure how soon I'll be able to do a good UniWB project, and I'm not even sure what the best things to shoot would be.

So, if anyone wants to do something, please do and post the results here!

Kirkt gave me a link to a good article by POTN member _GUI_ (Guillermo Luijken) on this for those who want to check it out:

http://www.guillermolu​ijk.com/tutorial/uniwb​/index_en.htm

In it he links to another article on Raw processing which needs to be translated (a utility is "built in" at the top of the page):

http://www.guillermolu​ijk.com/article/highli​ght/index.htm

Well, hey, I hope someone jumps on this! WhooHoo!