Harsh light - raw converter comparison

Started Aug 23, 2010 | Discussions
_sem_ Veteran Member • Posts: 5,033
Harsh light - raw converter comparison

Hi all,

here's an unscientific raw converter comparison regarding the conversion of a harsh-light image in a single conversion. There is a bit too much DR for the sensor to handle, therefore the specular highlights are slightly blown in the NEF.

The idea of the processing was to lift the shadows so that the skin detail would look approximately as bright and clear as it was at the scene, but not blow highlights more than they already are. Of course, no tuning can make the whole image look exactly like it was at the scene because the image DR is a few stops less.

Notice that Picasa is an outsider - its ultrafast raw converter has no tuning options, so the result is produced by using the Picasa fill-in light slider and a bit of saturation on top of its default conversion to JPG.

ViewNX does not allow much shadows lifting; eventhough the shadows are lifted less than with the rest, the noise level is higher, at the lowest sharpening level. The blown highlights (water reflections below) are best done by DxO I think.

Higher-res here: http://picasaweb.google.com/lh/photo/FjssEcP91krpSmDoWBLTn7k0FTC98Ldfw1DgDjMD5sA?feat=directlink (click the magnifier)

JEPH Contributing Member • Posts: 780
Re: Harsh light - raw converter comparison

Is something wrong with the ViewNX 1.5 highlight protection (if that's what the abbreviation means)? I think you've gone the opposite direction.

Have you tried the current version of VNX?

I'm thinking a lot of the qualitative results from conversions depend on users' experience with a particular converter. I'm not sure I'd agree with your assessment of VNX.

Thanks for posting. It's interesting. I've wanted to look at DxO but haven't yet.

OP _sem_ Veteran Member • Posts: 5,033
Re: Harsh light - raw converter comparison

JEPH wrote:

Is something wrong with the ViewNX 1.5 highlight protection (if that's what the abbreviation means)? I think you've gone the opposite direction.

I'd have to doublecheck, the labels could be misplaced, right now I just recall it did not do any good to this image. There is something "wrong": the name of the slider may be misleading. This slider seems to be the dual of the fill-light, it expands the highlights and upper midtones and compresses the lower midtones and shadows, quite useful sometimes and I miss it in DxO. But it is not about recovery of blown highlights. The latter is in VNX a matter of exposure compensation - but caring a lot for the (recoverable) highlights in VNX causes the shadows to suffer because there is a sort of a limited DR "window" that may be slided up or down with the EC slider but cannot be extended (unless one does multiple conversions and goes HDR).

Have you tried the current version of VNX?

Not yet. I hear the rendering has not changed (much?).

I'm thinking a lot of the qualitative results from conversions depend on users' experience with a particular converter. I'm not sure I'd agree with your assessment of VNX.

I'm sure it depends. I think I've tried all the sliders of VNX on this one, in many combinations.

Notice this is not a general assessment; this is specifically for harsh light and shadows lifting. If you take the same NEF and choose to make a high-contrast image with the skin reduced to a black shadow, VNX is fine. It does have certain other advantages.

Thanks for posting. It's interesting. I've wanted to look at DxO but haven't yet.

I think it is good for certain purposes. They tend to switch GUIs often

I've noticed Google does not do full-rez, here it is (click "Original" above image):

winparkman Veteran Member • Posts: 4,638
Re: Harsh light - raw converter comparison

Pulling things out of images is a mix of software and skill/experience. The more your software allows you to do, the more likely you are to recover your images. It is for that reason that I think you are better off moving away from free softwares that let you do a little to softwares that let you do a lot.
--
OK, not so purely a hobby.

 winparkman's gear list:winparkman's gear list
Fujifilm X100T Nikon D90 Nikon D700 Nikon D7100 Nikon AF-S Nikkor 24-120mm F4G ED VR +1 more
JEPH Contributing Member • Posts: 780
A problem

sem wrote:

JEPH wrote:

Is something wrong with the ViewNX 1.5 highlight protection (if that's what the abbreviation means)? I think you've gone the opposite direction.

I'd have to doublecheck, the labels could be misplaced, right now I just recall it did not do any good to this image. There is something "wrong": the name of the slider may be misleading. This slider seems to be the dual of the fill-light, it expands the highlights and upper midtones and compresses the lower midtones and shadows, quite useful sometimes and I miss it in DxO. But it is not about recovery of blown highlights.

No, you have a unique version of the software or you have made an error.

OP _sem_ Veteran Member • Posts: 5,033
Re: A problem

JEPH wrote:

sem wrote:

I'd have to doublecheck, the labels could be misplaced, right now I just recall it did not do any good to this image. There is something "wrong": the name of the slider may be misleading. This slider seems to be the dual of the fill-light, it expands the highlights and upper midtones and compresses the lower midtones and shadows, quite useful sometimes and I miss it in DxO. But it is not about recovery of blown highlights.

No, you have a unique version of the software or you have made an error.

So I checked, and I see I haven't misplaced the labels. The "shadows protection" "expands" the higlights around the reflections, "compresses" the clouds and darkens the skin. But I've also added some more shadows protection to lift the skin back. The lack of contrast somewhere is a typical consequence of using both shadows and highlights protection.

"wrong"... not really wrong, just the names of the sliders in different converters may be confusing.

winparkman wrote:

Pulling things out of images is a mix of software and skill/experience. The more your software allows you to do, the more likely you are to recover your images. It is for that reason that I think you are better off moving away from free softwares that let you do a little to softwares that let you do a lot.

Haven't checked myself but I hear the non-free CNX2 is just as VNX regarding this, as U-point does not allow local EC adjustment.

Regarding local adjustments, I've played with the brushes in LR3. However this was not as useful as I'd thought it'd be because one can't select all the highlights carefully while the available adjustments over the whole area didn't work well, except for the local contrast decrease.

winparkman Veteran Member • Posts: 4,638
Re: A problem

Brushes? I thought your problem was with the highlights and then, secondarily, skin color. You could use brushes but why wouldn't you adjust the light levels? What about Recovery? There are other things you could try before moving to brushes. Perhaps I am missing what you are trying to show.

sem wrote:

Haven't checked myself but I hear the non-free CNX2 is just as VNX regarding this, as U-point does not allow local EC adjustment.

Regarding local adjustments, I've played with the brushes in LR3. However this was not as useful as I'd thought it'd be because one can't select all the highlights carefully while the available adjustments over the whole area didn't work well, except for the local contrast decrease.

-- hide signature --

OK, not so purely a hobby.

 winparkman's gear list:winparkman's gear list
Fujifilm X100T Nikon D90 Nikon D700 Nikon D7100 Nikon AF-S Nikkor 24-120mm F4G ED VR +1 more
OP _sem_ Veteran Member • Posts: 5,033
Re: A problem

winparkman wrote:

Brushes? I thought your problem was with the highlights and then, secondarily, skin color. You could use brushes but why wouldn't you adjust the light levels? What about Recovery? There are other things you could try before moving to brushes. Perhaps I am missing what you are trying to show.

The demo is limited to getting the skin as bright as it approximately seemed to be at the scene, without getting to much noise and without blowing the higlights more than they were blown originally - in a single conversion and without local adjustments.
I didn't do anything regarding the skin colour; I did not adjust WB.

I did try other things, available in the converters... "recovery" is applied in the LR3 version; it does not apply in the others (except for the EC slider).

Generally I think brushes and other forms of localised adjustments (such as simulated grad ND, exposure masking etc) are valuable in this kind of work where heavy adjustments are applied and the inevitable DR compression always takes toll on the contrast somewhere. Some of these heavy adjustments would be better applied only where needed, not over the whole image uniformly. However this requires more pp work, and the support for it in raw converters is limited or none.

While test-driving LR3 I wanted to try also (not directly related to the shown demo) what local adjustments could be conveniently used on this image. Grad ND didn't work, due to the distribution of the highlights. I wanted to try something I've seen done using PS, dodging/burning with the tool set to highlights/shadows only so that the specular highlights stick out less (where one does not have to be very precise with the brush so it is a quick adjustment). Unfortunately, it does not seem to be possible to set the brush in LR3 to shadows or highlights only. Decreasing contrast locally did somewhat help, though not as efficiently as the way this was done in PS.

winparkman Veteran Member • Posts: 4,638
Re: A problem

Then I fail to see how this is much of a test of anything.

sem wrote:

winparkman wrote:

Brushes? I thought your problem was with the highlights and then, secondarily, skin color. You could use brushes but why wouldn't you adjust the light levels? What about Recovery? There are other things you could try before moving to brushes. Perhaps I am missing what you are trying to show.

The demo is limited to getting the skin as bright as it approximately seemed to be at the scene, without getting to much noise and without blowing the higlights more than they were blown originally - in a single conversion and without local adjustments.
I didn't do anything regarding the skin colour; I did not adjust WB.

I did try other things, available in the converters... "recovery" is applied in the LR3 version; it does not apply in the others (except for the EC slider).

Generally I think brushes and other forms of localised adjustments (such as simulated grad ND, exposure masking etc) are valuable in this kind of work where heavy adjustments are applied and the inevitable DR compression always takes toll on the contrast somewhere. Some of these heavy adjustments would be better applied only where needed, not over the whole image uniformly. However this requires more pp work, and the support for it in raw converters is limited or none.

While test-driving LR3 I wanted to try also (not directly related to the shown demo) what local adjustments could be conveniently used on this image. Grad ND didn't work, due to the distribution of the highlights. I wanted to try something I've seen done using PS, dodging/burning with the tool set to highlights/shadows only so that the specular highlights stick out less (where one does not have to be very precise with the brush so it is a quick adjustment). Unfortunately, it does not seem to be possible to set the brush in LR3 to shadows or highlights only. Decreasing contrast locally did somewhat help, though not as efficiently as the way this was done in PS.

-- hide signature --

OK, not so purely a hobby.

 winparkman's gear list:winparkman's gear list
Fujifilm X100T Nikon D90 Nikon D700 Nikon D7100 Nikon AF-S Nikkor 24-120mm F4G ED VR +1 more
OP _sem_ Veteran Member • Posts: 5,033
Re: A problem

winparkman wrote:

Then I fail to see how this is much of a test of anything.

I think you're being a bit cynical I didn't say it was something scientific or objective.

I've played with this some time ago, I thought it might be useful to somebody since harsh light & blown highlights threads tend to show up here regularly, and I thought I might pick up some useful comments.
http://forums.dpreview.com/forums/readflat.asp?forum=1034&message=36083625
http://forums.dpreview.com/forums/readflat.asp?forum=1034&message=35892832
http://forums.dpreview.com/forums/readflat.asp?forum=1034&message=35968798

winparkman Veteran Member • Posts: 4,638
Re: A problem

heh...OK, sem; you are right but I think the comments matched the test.

Mike.

sem wrote:

winparkman wrote:

Then I fail to see how this is much of a test of anything.

I think you're being a bit cynical I didn't say it was something scientific or objective.

I've played with this some time ago, I thought it might be useful to somebody since harsh light & blown highlights threads tend to show up here regularly, and I thought I might pick up some useful comments.
http://forums.dpreview.com/forums/readflat.asp?forum=1034&message=36083625
http://forums.dpreview.com/forums/readflat.asp?forum=1034&message=35892832
http://forums.dpreview.com/forums/readflat.asp?forum=1034&message=35968798

-- hide signature --

OK, not so purely a hobby.

 winparkman's gear list:winparkman's gear list
Fujifilm X100T Nikon D90 Nikon D700 Nikon D7100 Nikon AF-S Nikkor 24-120mm F4G ED VR +1 more
OP _sem_ Veteran Member • Posts: 5,033
Re: A problem

winparkman wrote:

... What about Recovery? ...

One more thing regarding Recovery in LR3, the hue shift of the blue sky towards greyish mentioned by some folks does show up in my sample. Unfortunately the beta has expired so I can't check how using different profiles might affect this (camera, untwisted, invariate, ColourChecker Passport profiles).

And the other thing regarding the recovery of partially blown highlights being generally considered better by NX than LR, however not showing in my sample. Firstly, the blown highlights in my sample are small and sharp-edged (a polariser was used), so the difference is not as easy to notice as in sunsets etc where NX was shown to produce less posterisation. Secondly, in order to make this show in my sample I'd have to apply negative EC, which would make things even worse in the lifted shadows. So I'd have to do either pseudo-HDR or high-contrast if I wanted to make use of this.

OP _sem_ Veteran Member • Posts: 5,033
Re: Harsh light - raw converter comparison

JEPH wrote:

Is something wrong with the ViewNX 1.5 highlight protection (if that's what the abbreviation means)? I think you've gone the opposite direction.

In fact VNX rescues highlights a bit better than DxO if you can afford to apply enough negative exposure compensation, on a different sample with gradual overexposure, like a blown part of the sky around the sun. In this sample the difference is practically too small to be noticeable even when pixel-peeping.

Keyboard shortcuts:
FForum MMy threads