Expose to the right?

E

eco_bach

Well-known member
Hi

I'm relatively new to the DSLR video world but thought it was just common sense to treat digital sensors, especially CMOS, exatly like reversal film in determining exposure.
ie always protect your highlites, in general you are better off underexposing a full stop to preserve highlite detail.

However a comment by Stu Maschwitz on Prolost.com and this following article has made me rethink this stratgey. Essentially the argument is that the lighter tones are able to record a greater number of distinct gradations, the more light, the more data, until of course you reach the sensor's saturation point and the highlites are clipped. Which means that a lot of your shots might appear overexposed. Although the article refers to RAW workflow only, I assume this would also apply to moviemode, ie in camera H.264. Shooting a couple test jpegs should be able to validate but can anyone else confirm, is exposing to the right(without clipping any data) always the better option?

http://www.luminous-landscape.com/tutorials/expose-right.shtml


and corresponding(long) thread on the subject, focus is on still processing but the same concepts apply. There is mention that ETTR is really only useful at 'base ISO's as well
http://www.luminous-landscape.com/forum/index.php?topic=49433.0

And another which simply shows the increased S/N with an overexposed image
http://schewephoto.com/ETTR/index.html
my takeaway from this being that you have more leeway with clipped highlites if shooting RAW, but otherwise, be especially careful you don't clip data if using ETTR.

And yet another must read for anyone contemplating creating their own Picture style curves

http://www.computer-darkroom.com/d30-profiling/d30_profile.htm
 
Last edited:
Y'know, the best answer to this is to test - I've been playing with the histogram sweet spot, but custom profiles seem to each need their own finesse. The "pancake" profile is the latest I've played with, and it can look washed-out & overexposed really easily - and pushing the darks back up seems to make the mids go pretty orange.
 
Agreed. Adding custom profiles to the mix really will alter the results. I'm using a beta of the default Marvels Cine profile. Need to conduct some simple tests, seeing how jpegs shot with different exposure settings hold up to gamma-exposure correction in post. Or simply correcting the resulting in-camera jpegs in Photoshop, and comparing the end results, may be all that's needed for a test.
 
The problem with this as a holistic approach is that it becomes easier to clip a single color channel without noticing if you're only relying on your camera's LCD/EVF and luminance histogram. With the GH2, for example, you can preview your RGB histograms in the camera's preview mode, so it may be better to take a test pic or clip first and evaluate it. If there's a lot of camera movement through different light ratios, this can get tricky. There is a technique for restoring clipped color channels in a program such as After Effects, but it's complex and doesn't always work.
 
ETTR is best for raw workflows as you have more room to grade in post. I don't know that it would work so well with H.264, as what you see is pretty much what you get. An ETTR image will often look flat and washed out but will retain better detail with less noise in the shadows. Correcting for best image quality requires being able to tweak the gamma curve and levels in ways encoded video is not very amenable to.
 
Exposing for your shadows is typically the best plan if grading in post is part of your workflow. The logic behind this is described thoroughly in the urls you linked, and there's not much of a counter-argument. The only time I let my shadows clip is if I'm trying to get the look in-camera. This has happened to me on a number of occasions when there simply is not enough time to do any color work in post, so it is best to let shadows be shadows, even if they're not NTSC legal. Also, contrast and saturation are typically added in a post workflow, so if your shadows are clipped, they stay clipped, and the legal shadows just get darker, disguising the data loss in the clipped shadows. But if your highlights are blown out, and you try to add contrast, saturation, etc., or if you try to mess with gamma curves, you may see a milky, flat, pasty area rise from the blown-out portion of your image, as darkening those areas just makes them more apparently gray because there is no data available.

Short version: If color grading is budgeted in your workflow, expose for your shadows (aka expose to the right). If you need to get the look in-camera and/or you don't have time for any color in post, expose as safely on the histogram as possible, but favor keeping your highlights safe rather than your shadows.

This is all IMHO.
 
You must log in or register to reply here.
Back
Top