Linear response curve in C9
Hi,
after upgrading to C9 I'm still a bit hesitant in upgrading my files from 8.3.4 so for now I'm playing a bit around.
I've already discovered the power of the Curves tool used via Local Adjustment - especially for those typical november/december shots with sky highlights blowing pretty easily while the shadows are dark and deep. A gradient mask with some local fine edit with Erase brush and Paint brush followed by a suited curve and, "ta-da", a very believeable, natural looking lifting of the shadows.
But this testing took me to the main question, the Linear response curve - nothing particular new in C9, I know, but up to now I've never got deeper into this.
Now it was almost a "Wow-experience" when I switched an image (person's face more or less close-up) from Film Standard to Linear response since the skin tones looked to me so much more natural. Ok, the lighting got pretty dull - as expected - and certainly needs a different approach but the skin tone (same WB setting) as well as the overall tone graduation looked much more natural to me while the version based on Film Standard seems to look somehow harsh. What I also noticed is that the Linear version needed a lot more exposure adjustment to bring it to the right brightness levels - but this doesn't seem to affect the noise levels as it usually does when greater amounts of exposure correction are applied.
Thoughts?
after upgrading to C9 I'm still a bit hesitant in upgrading my files from 8.3.4 so for now I'm playing a bit around.
I've already discovered the power of the Curves tool used via Local Adjustment - especially for those typical november/december shots with sky highlights blowing pretty easily while the shadows are dark and deep. A gradient mask with some local fine edit with Erase brush and Paint brush followed by a suited curve and, "ta-da", a very believeable, natural looking lifting of the shadows.
But this testing took me to the main question, the Linear response curve - nothing particular new in C9, I know, but up to now I've never got deeper into this.
Now it was almost a "Wow-experience" when I switched an image (person's face more or less close-up) from Film Standard to Linear response since the skin tones looked to me so much more natural. Ok, the lighting got pretty dull - as expected - and certainly needs a different approach but the skin tone (same WB setting) as well as the overall tone graduation looked much more natural to me while the version based on Film Standard seems to look somehow harsh. What I also noticed is that the Linear version needed a lot more exposure adjustment to bring it to the right brightness levels - but this doesn't seem to affect the noise levels as it usually does when greater amounts of exposure correction are applied.
Thoughts?
0
-
I would say that
(1) there is no point in updating all your existing images to the new processing engine. If you have achieved the look you want on the previous version, you may find that changing to the new engine alters the appearance of the image - for instance contrast, brightness and saturation work a bit differently, and you might then need to make further adjustments. On the other hand for images that you want to make improvements to, the new engine would offer more opportunities.
(2) My understanding of the different response curves is that if you select linear response in the base characteristics, you could (in theory! - but it might be hard work) adjust the image until it looked like it did with Film Standard, and if you started with Film Standard, you could (in theory) adjust it until it looked like it did with the linear response. All the options apart from linear apply a curve before you start, to give you a starting point on top of which you make your adjustments. If you find it easier to get a result you are pleased with using linear for some images, go for it. They do look duller out of the box, but I am not sure that different exposure adjustment is necessarily the thing. If you change an image to Linear, it looks duller, but blown out highlights will still be blown out - the top end of the histogram is still climbing the right hand edge. The brightness and curves tools are more likely to be the way to go, I suggest. (Just my amateur opinion, though.)
Ian0 -
Neither of the curves (linear or film standard) has a really neutral appearance, and there would be room for improvement regarding an optimum starting point. Standard film curve pulls down the darker tones too much and does affect the natural appearance and obscures some of the finer gradations and color and tonal details. So I switched over to linear and made some curve presets to preserve more of the tonal and color details without having the flat washed out appearance of linear default. But I noticed with especially bright blue skies, that with larger exposure changes, also the blue color shifts and changes totally in areas that are close to the brightest values (but not clipped), something that absolutely does not happen with the standard film curve however much you push or pull. So something is not quite right with the linear curve and colors. I stopped using it, and now start with the extra shadows curve, which offers a more realistic starting point to my taste than the standard film curve, which is in need of a proper revision, as it is simply overdone, and does no justice to todays high dymamic range, color depth and resolution cameras. On a side note: CO1 v9 is a very nice update, and with the shadows/highlights tool update from v8, the new contrast/brightness/saturation tools give fantastic results. Together with a set of revised and improved (film) curves it could even be much better still. Perhaps a nice upgrade feature for 9.1?
Chris0 -
[quote="Michael11" wrote:
Thoughts?
first of all you need to visit a great topic /yes, it takes time to read through it/ @ LuLa = http://forum.luminous-landscape.com/ind ... c=100015.0 - where you can read about C1 curves (not from the curve tool, mind you) and also
http://www.ludd.ltu.se/~torger/dcamprof.html
http://www.ludd.ltu.se/~torger/photogra ... iling.html
it is a great tool to create profiles for C1 nowadays (you can still use argyll alone, rawdigger + argyll or commercial software like ProfileMaker - but DCamProf has the best set of options, and it can be combined with rawdigger too if you want) _and_ the best manual/tutorial vs anything else out there)
briefly
1) C1 starts (I mean before any of your adjustments through C1's user interface) its work with raw by applying a certain curve (let us call it "transfer function") behind the scenes (you do not control that, it is hardcoded) to raw data + their profiles have LUT that also applies some "curves" effectively (they increase brightness that way for example) - so using "linear" curve (or "linear scientific", which you can appropriate from CH files even w/o CH license - just rename and copy, with regular profiles ) does not make things "linear" - you need to create your own profiles compensating that "transfer function"... see above mentioned URLs...
2) C1 curve files (.fcrv) might have instruction inside to C1 code to provide blown/clipped (in raw rgb data that C1 is working with) areas reconstruction - you can easily see this comparing "linear" and "linear scientific" curves ... "linear scientific" is intended to be used when you do not have any clipping in raw and when you do not expose close to any possibly non linearities near clipping in raw channels (some camera have that behavior when manufacturer tries to get as much DR as possible, etc)
3) P1's own profile supplied with C1 (excluding possibly reproduction profiles for CH edition for certain P1 cameras) were created for non ETTR type of shots, so they increase the brightness for less than ideally exposed shot and make an ideally exposed shot in most cases too bright (so you have to pull exposure slider for examle)... as we all know C1 still can't paint overexposed/clipped (in raw) areas properly like some other converters can - specifically providing pleasant transitions from clipped areas to non clipped in many if not most cases... it is a shame now that we already on v9... so with C1 you certainly either create your own profiles and expose ETTR-like with "linear" curve (or create profiles properly compensating for "transfer function" + the curve intended to be used instead of "linear", DCamProf has the manual/tutorial with instructions) or stay away from ETTR-like shots if you like to use OEM profiles and "film" curves, etc... and unlike with ACR/LR I'd rather stay away from clipping at all - again how C1 paints clipped area transitions (w/o heave local editing) is below competitions... my $0.020 -
[quote="deejjjaaaa" wrote:
[quote="Michael11" wrote:
Thoughts?
first of all you need to visit a great topic /yes, it takes time to read through it/ @ LuLa = http://forum.luminous-landscape.com/ind ... c=100015.0 - where you can read about C1 curves (not from the curve tool, mind you) and also
http://www.ludd.ltu.se/~torger/dcamprof.html
http://www.ludd.ltu.se/~torger/photogra ... iling.html
it is a great tool to create profiles for C1 nowadays (you can still use argyll alone, rawdigger + argyll or commercial software like ProfileMaker - but DCamProf has the best set of options, and it can be combined with rawdigger too if you want) _and_ the best manual/tutorial vs anything else out there)
briefly
1) C1 starts (I mean before any of your adjustments through C1's user interface) its work with raw by applying a certain curve (let us call it "transfer function") behind the scenes (you do not control that, it is hardcoded) to raw data + their profiles have LUT that also applies some "curves" effectively (they increase brightness that way for example) - so using "linear" curve (or "linear scientific", which you can appropriate from CH files even w/o CH license - just rename and copy, with regular profiles ) does not make things "linear" - you need to create your own profiles compensating that "transfer function"... see above mentioned URLs...
Although I have not yet delved into the hidden workings of CO1 for the sake of custom profile making, I can follow your points. Even with the "linear" curve, CO1 blows highlights with certain images at default settings, that do in fact not really have any blown highlights. This can be seen with e.g. Raw digger or Fast raw viewer, but also with DxO optics, that shows the same images at default settings, with absolutely no highlights blown or even near blown. So also with the linear curve, CO1 boosts the highlights by a significant margin. although the image looks flat and underexposed. This seems to indicate, that an exposure boost is applied before any choice of curve or corrections, and choosing the "linear" curve, CO1 actually applies a second curve to make the image appear linear, but the blown highlights remain. For this reason I have stopped using the linear curve. Instead I underexpose slightly, and choose the extra shadow curve and go from there. This prevents all sorts of issues with clipping and color shifts in the highlights. It may well be that Sony (I use the Sony A7r) also does something to their raws, but that does not neccessarily give bad highlight roll off or DR, because in DxO it is pretty hard to blow the highlights at no-correction setting (switching off auto correction).
DxO has a pretty integer treatment of the raw files it seems: they apply a fairly linear-like curve, preserving lots of tonal gradations and maximizing DR, as a default, and on top op that, you can choose their "generic" renderings, with a default setting of your detected camera body and linked curve/color correction. You can slide in the amount of correction from 0-200%. At 0% you are back to their underlying "neutral colors, neutral tonality" linear-like curve. This is a superb implementation, which I would love to see with CO1: start with a truely linear-like curve and color rendition, that maximizes tonal detail and DR and gives reliable neutral colors, and then allow the option to "slide in" their own interpretation of what your camera body should look like. In DxO this works superbly: I use 35%-50% of the camera body preset and get spot on colors and a large DR. The neccessary contrast to prevent a flat look from the linear like curve is applied as part of the camera body preset, so no need to fiddle around with the curve tool in 90% of the shots.
In CO1, I have developed a routine to get pretty accurate colors: I shoot a X-rite color target in bright Sunny daylight, apply a white balance correction, and then correct the colors by eye with the color target and then a suitable set of shots, keeping a window open with DxO optics at no correction setting to correct the colors with the advanced color editor, taking some ten correction settings usually. It may sound unreliable, but I have been amazed at how much better the corrected profile has come out for my Sony and Pentax cameras.[quote="deejjjaaaa" wrote:
2) C1 curve files (.fcrv) might have instruction inside to C1 code to provide blown/clipped (in raw rgb data that C1 is working with) areas reconstruction - you can easily see this comparing "linear" and "linear scientific" curves ... "linear scientific" is intended to be used when you do not have any clipping in raw and when you do not expose close to any possibly non linearities near clipping in raw channels (some camera have that behavior when manufacturer tries to get as much DR as possible, etc)
3) P1's own profile supplied with C1 (excluding possibly reproduction profiles for CH edition for certain P1 cameras) were created for non ETTR type of shots, so they increase the brightness for less than ideally exposed shot and make an ideally exposed shot in most cases too bright (so you have to pull exposure slider for examle)... as we all know C1 still can't paint overexposed/clipped (in raw) areas properly like some other converters can - specifically providing pleasant transitions from clipped areas to non clipped in many if not most cases... it is a shame now that we already on v9... so with C1 you certainly either create your own profiles and expose ETTR-like with "linear" curve (or create profiles properly compensating for "transfer function" + the curve intended to be used instead of "linear", DCamProf has the manual/tutorial with instructions) or stay away from ETTR-like shots if you like to use OEM profiles and "film" curves, etc... and unlike with ACR/LR I'd rather stay away from clipping at all - again how C1 paints clipped area transitions (w/o heave local editing) is below competitions... my $0.020
Post is closed for comments.
Comments
4 comments