Pink Areas on Over Exposed

[r0man]

Similar to afx I dont see such drastic changes with Olympus cameras. HR does cause range compression so there is slight reduction in contrast and saturation, but thats the same for the other software I've got. The range slider does work for me with 0 effecting the whole histogram and 100 a stop or two below the white point.
There's slight difference in the images I posted above, and the afterHR image needed +5 contract and +20 vibrance to look pretty much the same as the original.

Oh, it would be nice to have saturation/vibrance sliders in the exposure tab as well as the contrast for this reason.

[KeithR]

Hmmm, two steps forward, one step back.

<snip>

The HR Range slider doesn't seem to affect this at all, and like I said, I can't control the magnitude of the problem at because just going from zero to one is such a huge step.

<snip>

Am I doing something wrong? This seems to be a significant problem. I can compensate for the contrast/saturation problems after fixing highlights, of course, but why is this effect present?

Well I must be doing the same thing, because I get the same unacceptable results on my Canon 7D files, and no "recovery" to speak of, just a flattening of the highlights - and the range slider makes no observable difference whatsoever.

Lr 4 and RawTherapee highlight recovery tools actually recover detail from the same files.

[gregglee]

HOWEVER... just moving the Highlights slider from zero to one seems to have a significant detrimental impact across the entire photo frame. It shifts everything to the left on the histogram, quite a bit, reducing contrast and saturation on everything. Blue skies become grayish-blue, and green grass becomes grayish-green. Even in midtones, where I'd expect HR to have minimal impact.?

While I don't see such a great effect, on most images I do see a big jump between HL = 0 and HL=1 with impact well down from what I consider "highlights." Perhaps the new HL range slider is intended to adjust how far down the histogram HL acts. But I haven't seen that it has much effect. Help still discusses the monochromatic check box, which was often useful, but is now is gone, and does not discuss the range slider.

The big first step in HL is why I often use exposure (plus fill or bez) to reduce highlights if I can. Usually no small adjustment is possible with HLR. while there is using with exposure.

I hadn't thought about it but first step in HLR reminds me of the broken fine adjustment of clip levels in autolevel and autocontrast. But at least HLR adjusts smoothly for step 2 and higher.

Edit: To be clear the big first step in HL is not new to this release.

[greerd]

This happens to me somewhat with my .rw2 Panasonic files. Going from 0 to 1 moves the histogram to the left much more than I would like and it takes some of the blue out of the sky leaving it more grayish-blue. I don't notice any loss of contrast or gray veil over the rest of the image though.

[Quicksand]

It would help if you would stated on which files this happens.
No such effects here on my Olympus and Nikon files. There is an increased reduction of contrast when pulling up the HR slider, but nothing as drastic as you describe by just activating it.

Uh, right! I guess that would help, wouldn't it.

Pentax K-5. That's me over here.

Anyone is welcome to download a sample raw file (the sign-and-fence shot from that linked Bibble bug post) if you want to take a shot at it. This is the file I have returned to, again and again, to see if highlight recovery is fixed.

I have plenty of others, but this one shows the "gray veil" effect on the sky and the grass.

[afx]

Hmm,
No Grey veil here, just a slight drop in contrast when going from HR 0 to HR 1. Checked both on Linux and Windows on calibrated screens.
Is there any other setting active on that image on your box?
The Histogram drop looks quite pronounced though. More than on other cams I checked, but similar to other PEFs.

cheers
afx

[grubernd]

the histogram drop is different on different cameras. confirmed.
but i havent seen any adverse effects - and i have a wild mix of cameras in usage.
nikon D700, D80, panasonic G2, canon G11 -- those are just the current ones.

as a sidenote: nikon is still the only camera manufacturer i know of that is able to create raw-files that gracefully transition from highlights to "nothing". all others produce harsh edges. just in case anyone is wondering why nikon-users complain less about highlight problems. it's a better raw, really, because nikon doesnt leave it raw..

anyway, back to the HR stuff, here's the important stuff you need to have in mind:

going from 0 to 1 and beyond really switches on the HR algorithm.
0 is no Highlight Recovery.
1+ is with Highlight Recovery.

a UI-design decicion i personally dislike, because it veils what's happening.
at 0 it is not HR at a setting of 0 but HR not activated, which is something different.

so think about going from 0 to 1 as if you'd tick a huge big ON-box instead of changing a setting.
with that in mind, HR works very much as expected.

[andysalay]

a UI-design decicion i personally dislike, because it veils what's happening.
at 0 it is not HR at a setting of 0 but HR not activated, which is something different.

so think about going from 0 to 1 as if you'd tick a huge big ON-box instead of changing a setting.
with that in mind, HR works very much as expected.

+1000 to bold highlighted statement.
I was complaining already when the change was introduced. Not logical, and is forcing me to use doubleclick on the label to zero, and then CTRL+Z to restore the value, if I want to toggle the effect for comparison purposes.

Needless to say, when you forgot the (applies also to the other) option zeroed, and quit ASP session (working in FS mode), there is no way to recall the original value.
The exact reason, why the checkboxes were so useful.
Unfortunately, somebody has seen it other way.

Regards,
Andy Salay

[maggus]

Based on a very small sample, the problem may be less with the new version, but certainly did not completely eliminate

If you find an image that still produced magenta or colored highlights, please upload a sample to us.

Hi Jeff,

good and bad news ....
Today I found some time to roughly check HR issues on EOS 60D and found some images where it got better ... some still same "bad" level, and some got worse compared to V1.0.0-39.

I uploaded the JPG output to THIS galery.

• Images processed in ASP 1.0.0-39 and 1.0.1-10 (ref. also EXIF data).
• Images with some text are from V1.0.1-10, the textless images are from V1.0.0-39.
• RAW opened with defaults, but NN and sharpening disabled.
• Only Highlight recovery adjusted on the images, nothing else (OK, on the tree I also adjusted the blacks slider a bit).
• No tricks about BEZ or exposure applied.

Will start uploading following images to your upload-link soon:

• img_2092.cr2
• img_4676.cr2

since those are the "bad" once.

Just PM me if you also want the other RAWs.

cheers
Markus

Update:

• upload finished, ID 300 + 301

[gregglee]

going from 0 to 1 and beyond really switches on the HR algorithm.
0 is no Highlight Recovery.
1+ is with Highlight Recovery.

a UI-design decicion i personally dislike, because it veils what's happening.
at 0 it is not HR at a setting of 0 but HR not activated, which is something different.

so think about going from 0 to 1 as if you'd tick a huge big ON-box instead of changing a setting.
with that in mind, HR works very much as expected.

I think the check box died with Bibble 4. Bibble 5 doesn't have it.
I had also concluded that in B5 and ASP, 0/1 is the on/off switch.

But this is not the point. However the interface works, HL from 0 to 1 in Bibble5 was normally a small change. In ASP it is normally a large one that skips over the Goldilocks zone. Overshhot by HL must be offset by raising exposure or on curves pushing the right slider left. Also requires pushing gamma slider left to reverse the flat appearance that others have mentioned.

So either HL was recalibrated to make the HL=1 stronger. or maybe this is another unintended consequence of the processing order changes that have been cited as the origin of the broken autolevel/autocontrast/clip levels.

Whether this is a direct contributor to pinks spots is unclear, but it will certainly be in play for any changes to HL tool.

[Jeff Stephens]

Just a quick note: thanks for all the sample images - these are VERY useful. We'll be downloading all of these, combining them with our current HR test set and this analyze the results and see if there is an underlying issue in our HR or if it can be improved. Uploading samples to us at http://aftershotpro.com/upload is always a bit easier that posting your own links (because we control that upload system, the attachments are automatially mirrored locally, and we can cross-reference them later, if needed), but if you've provided images through other services, don't bother re-uploading them using our upload facility at this point...

Cheers, Jeff

[maggus]

Hi Jeff,

I guess you know ... whenever you need some samples, just ask me/us ... I´m sure you´ll get them.

cheers
Markus

[simico]

Anyone is welcome to download a sample raw file (the sign-and-fence shot from that linked Bibble bug post) if you want to take a shot at it. This is the file I have returned to, again and again, to see if highlight recovery is fixed.

I have plenty of others, but this one shows the "gray veil" effect on the sky and the grass.

I downloaded the pef and noticed that:

• highlight recovery didn't recover anything
• it just pushed (quite a big push) most of the histogram to the left, it's like lowering exposure without affecting the blacks
• color of the sky and grass changed

Here's how it looks.
Honestly, what I can see with this pef is not highlight recovery at all, it's just lowering exposure without touching the shadows.

[afx]

Honestly, what I can see with this pef is not highlight recovery at all, it's just lowering exposure without touching the shadows.

What else would you see if there are no details in the highlights?
If there where cloud structures or a bridal dress that would be more relevant.

cheers
afx

[Denis de Gannes]

Also the image foreground is in focus, so the background which is far in the distance is out of focus so there is no detail to recover.