"Exposing to the right" is over exposed, what now?
in [Photoshop]
|
From: Tech Play on 5 Oct 2009 23:11 On Mon, 05 Oct 2009 21:18:08 -0500, Doug McDonald <mcdonald(a)NoSpAmscs.uiuc.edu> wrote: >Floyd L. Davidson wrote: > >> But regardless of that, the cited URL above from >> luminous-landscape is *not* full of good stuff. They >> miss the point entirely, and provide nothing that is >> actually useful! All that changing in-camera contrast >> does is compress/uncompress the histograms idea of >> middle gray. It does not change where the highlights >> fall, and thus does not change what you would set for >> the exposure. (Does anyone ever look at the shape of an >> in-camera histogram and make adjustments as a result? >> It's a meaningless excercise, which will not change the >> camera raw data. The shape of a post processing >> histogram is very useful, but not an in-camera >> histogram.) >> > >What you say is wrong for the Canon 30D. The histogram does >not change the raw data stored by the camera. But the camera setting DOES >change the camera displayed histogram. IF you set the camera "comtrast" >setting high, the highlights in the histogram get pushed way up >and are indeed useless for setting exposure. Ii you set it low, >at -3 or -4, the histogram displayed by the camera agrees quite nicely >with how close you are getting to clipping. At -4 the blinking clip display, >the histogram, and the raw file all agree well about how close you are >to clipping. > >And yes, I do look at the in-camera histogram. It works. > >Doug McDonald If you had a CHDK capable camera, you could use any of the various live-view RGB color histograms to find out which of any color channels need to have their contrast reduced too. I found that an overall -2 on contrast on the "custom color" settings along with a -2 setting on the R channel of Canon cameras gives a more accurate rendition, provides maximum dynamic range, and doesn't clip the R channel first on highlights. I could reduce B and G by a -2 also for even more dynamic range, but then I'd lose out the nice balance. And, quite frankly, it already has more dynamic range than any film I used to use, so it's not all that important. Few if any interesting and well composed shots will benefit from more dynamic range. A large dynamic range is the crutch of snapshooters that don't know how to properly expose a subject in the first place and will then always try to depend on post-processing to fix all their beginner's errors.
From: D. Peter Maus on 6 Oct 2009 03:51 On 10/5/09 16:44 , Charles wrote: >>> Porte >> For me, digital is the opposite of film in exposure emphasis. >> In film, you expose for the shadows, while in digital you should expose >> for the highlights. >> You can't print underexposed shadows on a film negative but you can burn >> in overexposed highlights since the negatives tend to hold some >> information in that area. >> You can't print overexposed highlights in a digital image but you can >> tease out information from underexposed shadows in digital processing. >> You can't treat your entire image the same way, so you need to apply >> techniques similar to dodging and burning in the digital realm using area >> adjustments of an image and the various tools/techniques available in >> image editors. > > Exactly! The clipping that occurs in digital photography is rather harsh > ... all scene details more luminous than the clipping (saturation of the > sensor) point are gone, gone, gone. You can't tease them out with any > amount of post-processing, regardless of how heroic or exotic. I fail to > understand why such a basic concept is so hard to understand. > You've answered your own questions, here. > Film does not "clip" in the same manner and some folks just don't seem to > get it. There are curves with soft knees and then there are curves with > very sharp knees. They are very different (in terms of useful output) when > the input data approaches and exceeds them. A soft knee means that there is > compression of the data and that means that a correction that employs > decompression can restore the data, at least to some extent. > > So, expose to the right if you can afford lost high-lites ... and I do mean > to emphasize LOST. Like a hair-cut, clipped is gone. > Not all highlights contain details. These can be lost. That reflection in a Packard's bumper on a bright, sunny day is just a bright spot. Expose to retain that detail and you'll be looking at the kind of night shot done by Hollywood in the days of "Yancy Derringer." Accept that those specular reflections contains no detail, and you can easily expose more 'to the right' without any loss of image detail, while recovering more low light detail with lower noise. Dynamic range compression with film was a limitation that shooters learned to work around, embrace, or simply accept. And like tape hiss in analog audio, it covered a multitude of sins. The linear nature of digital is no different. It's a limitation that requires practical evasions. You either learn to work with it, or not. And the output reflects your choice.
From: Paul Furman on 6 Oct 2009 10:24 Floyd L. Davidson wrote: > Doug McDonald <mcdonald(a)NoSpAmscs.uiuc.edu> wrote: >> Floyd L. Davidson wrote: >> >>> But regardless of that, the cited URL above from >>> luminous-landscape is *not* full of good stuff. They >>> miss the point entirely, and provide nothing that is >>> actually useful! All that changing in-camera contrast >>> does is compress/uncompress the histograms idea of >>> middle gray. It does not change where the highlights >>> fall, and thus does not change what you would set for >>> the exposure. (Does anyone ever look at the shape of an >>> in-camera histogram and make adjustments as a result? >>> It's a meaningless excercise, which will not change the >>> camera raw data. The shape of a post processing >>> histogram is very useful, but not an in-camera >>> histogram.) >>> >> What you say is wrong for the Canon 30D. The histogram does >> not change the raw data stored by the camera. But the camera setting DOES >> change the camera displayed histogram. IF you set the camera "comtrast" >> setting high, the highlights in the histogram get pushed way up >> and are indeed useless for setting exposure. Ii you set it low, >> at -3 or -4, the histogram displayed by the camera agrees quite nicely >> with how close you are getting to clipping. > > The brightest part will still show at *exactly* the same > place. Maybe for contrast (though I doubt it) but WB makes a big difference. I just tested at 25K & 100K, the highlights are completely different. I'm looking at split RGB histograms btw. > If you set exposure so that the brightest part > of the histogram is just barely less than the right edge > of the graph, it will be the same regardless of how > in-camera contrast is set. What contrast will change is > how much of the area of the graph is spread out over the > entire graph (low contrast) as opposed to being peaked > in some portion of the graph (high contrast). > > Changing the contrast does *not* change the value of the > brightest part of the image, which is what you use to set > exposure by the "expose to the right" method. > >> At -4 the blinking clip display, >> the histogram, and the raw file all agree well about how close you are >> to clipping. > > If you want to see a good demonstration of how it works, > set your camera to maximum contrast, use manual > exposure, and take a picture of an indoor light fixture. > It will be dramatic if the area around the light fixture > is relatively white and there is little light from any > source other than the fixture. Zoom in or be at a > distance where the light fixture is a significant part > of the image area, but most of it is the area around the > light and thus darker. > > Look at the histogram, and experiment with exposure. > The idea is to find the exact exposure where the > histogram bumps the right side of the chart. More > exposure will cause a spike to climb up the right side, > less exposure will cause gap to form between the graph > and the right side. Perhaps the ideal exposure for this > experiment is where there is just a small spike, because it > will be easiest to compare the height of the spike to see > if exposure changes. > > Once you have that, adjust the in-camera contrast > setting from maximum to minimum, and check the histogram > for each exposure. What you'll find is that the little > spike at the right side doesn't change. What does > change is the spread of the less bright areas of the > image. If you shoot a bare bulb and have a dark > background the entire bulk of the graph will be over on > the left side. It will be wider with lower contrast and > narrower for high contrast. The amount of area will > depend on how closely you zoom in on the light fixture > (the ratio of light source to background), and the > spread that it has along the right side of the histogram > will depend on the difference in brightness of the light > and the background. > > Adjust the physical attributes to get a useful > histogram. Then change the contrast setting to see what > it actually does affect. > >> And yes, I do look at the in-camera histogram. It works. > > But what do you look for? Not the distribution of the > area under the graph! That's what the contrast setting > will change. It does not change the highest value of > the displayed graph though (which exposure does change). > -- Paul Furman www.edgehill.net www.baynatives.com all google groups messages filtered due to spam
From: Charles on 6 Oct 2009 18:27 > Digital clipping is almost never "saturation of the > sensor". Unless you are shooting at the base ISO > clipping is saturation of the Analog-Digital-Converter > (ADC) by sensor output that is greater than the maximum > ADC input. And what difference would that make? There is no sharper knee than one imposed by basic mathematics. 2^8 = 256, is a simple example. With 8 bits, the maximum value is 255, base 10, with NO exceptions possible. You actually help to make the case that over-exposed digital shots guarantee lost information in the bright portions of the scene. I continue to believe that exposing to the right is perpetuated by folks with limited experience and knowledge. They are perhaps into the dark parts of scenes and don't understand dynamic range. Many of us are more greedy and want details at both ends of the luminosity range. Hey, blow those high-lites out! If that is the intended message of a given scene, go for it. I don't care for over-exposed shots ... they are just not my thing. I have seen some high-key shots that were attractive and effective and do appreciate the skill and art of the photographers who can pull that off. Specular high-lites might deserve to be blown out. Not the point of my response. I'll just say it one more time and then go away ... if it is clipped (by a sharp knee response curve ... due to sensor saturation or A/D saturation or any saturation link in the chain) it is GONE!
From: mike on 6 Oct 2009 18:37
In article <hadpc1$8up$1(a)news.eternal-september.org>, charlesschuler(a)comcast.net says... > > So, expose to the right if you can afford lost high-lites ... and I do mean > to emphasize LOST. Like a hair-cut, clipped is gone. > But unlike a hair-cut it never grows back. More of an amputation methinks... Mike |