Hey Martin,
I have another question for you, since I’m just beginning to learn in the ins and outs of gamma curves:
I noticed in the original Marvel’s Cine Flat Profile 1.2, the gif of the gamma curve you posted seems to have a fairly pronounced toe, but not so much of a shoulder. This looks like to me that you’re pushing the apparent middle part of dynamic range a bit higher up. And I’ve noticed in my tests that the picture occasionally looks a tad dark if I expose for it correctly with a incident meter light meter (especially in low-contrast scenes.)
Should I be compensating +1/3 or +2/3 a stop for proper exposure with this profile? Or should I be exposing normally (with a light meter) and adjusting the picture to my preference in post? Or should I just not use a light meter and eyeball it instead?
Am I misunderstanding the gamma curve?
-Dave
Hello Dave.
What really is going on, is that we apply a curve on a curve. The sampling, filtering and further post-processing of the camera’s image-sensor is a very complex matter and we can’t imagine (and are not being told) what goes on “behind the scenes”. You can get an insight into the complexity by reading this wikipedia link on the Bayer Filter. After the camera eventually has rendered an electronic representation of a real world image, many extra circuits – both analog and digital – are put to work to turn it into anything useable and digital. With the extra handicap that the video image is further processed and compressed.
The picture profile settings (and a few extra image settings that can be accessed through the menus) are just the tip of the iceberg of the motherload of parameters and circuits that can be tweaked. Not only “simple” electronics are involved, but quite a lot of physics as well.
The eventual image (based on the selected preset profile, e.g. Neutral or Faithful) already has a quite complex curve applied to it electronically sometime during the processing described above. In fact, there already is something in place that resembles an S-curve! What we do by altering the curve, making that existing curve more pronounced. It also means, that when we pull it too much, relations between color channels are disturbed, resulting in “plastic skintones” and other weird and unexpected artifacts.
To get back to your first observation. The Canon D cameras all seem to underexpose when using an external light-meter, and this seems to have to do with the way the camera handles the ISO setting. Specially when you are using a flat profile, you can push it much more, maybe indeed to +2/3. The built in histogram display is your friend here! Try to Google on the subject and you will find many discussions regarding this issue. Some propose to lower the ISO setting on the meter.
You are right about the curve. It is S-shaped and you can see a typical picture of the Marvels cine profile below (many people have asked for this, so here it is!). Because it’s applied on top of an existing curve (we have no other option, we have no Raw), the actual resulting curve is hard to calculate/visualise without reliable measurement. Digital video cameras such as the Sony EX-1 can deliver a number of test signals to the post-processing electronics, so the effect of curves and painting can easily be analysed. The Canons don’t have this option. What one could do, is shooting calibrated multi-level grayscales and analyse the results. A graph can be drawn from those results, most probably resulting in a wobbly s-curve.

Click for enlargement
Also, changing the profile’s Contrast setting alters the curve – it is NOT linear. So, having said all this; using your eyes and the histogram is still your best option, and getting some experience with the profiles and the resulting footage.
I am currently using the new Magic Lantern firmware patch that gives me a spot-reading, so i can e.g. expose human skin on 75% and check highlights. This and watching the histogram gives me good results.
I wish you all a happy and prosperous 2011!
Martin Beek.
Recent comments