An elder architect hammered us in UNI with every single line classification, "rectilinear" gosh (or literally straightlinear), "curvilinear", etc, really useful for types of lens perspective projections, i'm sure colour science could benefit from that mouthful
The definition I like using for Scene Linear is this: It's a space where the code value of the pixel is proportional to the quantity of photons (IE energy) that hit that pixel. The implications for this aren't necessarily obvious. Anyone who remembers their intro to photography class will recall that a one-stop increase in exposure means you're doubling the amount of light. Thus, you have to double the amount of photons caught by the sensor and that's a 2x gain to your scene linear state image. Suppose you wanted to simulate the effect of a single color light leak in the camera. You may have a scene linear image of the scene (with no light leak) and in addition, you also want to expose the sensor to the light of the light leak. In linear, you're just going to take the scene linear image of the scene and add on the color of the light leak (times whatever exposure you want it to be) and that's your result. Suppose you shot the scene and you're wondering how it would look if you had used a blue filter. Blue filters transmit mostly blue light and block most of the red and green light. Suppose you want to recreate a filter that blocks 70% of green and red photons. You can then multiply your scene linear image blue channel by 1.0x and multiply red and green by 0.3x, and now you've got a rudimentary guess of how it would have looked. There are so many different things you can do in linear, and they're fairly easy to figure out by just thinking about what the operation is doing from a physics standpoint.
A couple further notes. By "proportional" I mean suppose your scene is a grey card with X watts of light bouncing off of it. Ultimately, you photograph it and bring the image into scene linear and measure a code value Y for all channels. You should find that there's some value k for which Y = k*X for every pixel in the shot. My definition of linear differs from log images. The definition of log would instead be "an image where the code value of a pixel is proportional to the logarithm of the amount of photons that hit that pixel", though in practice our log encodings have a few more terms that it's more just "roughly proportional" at this time.
in scene linear, wouldn't you use offset to add light that wasn't there? you'd use gain to increase light, but offset to add light that wasn't already present right?
@@factarchive9576 Pretty much. Adding an offset in linear is analogous to if you shot through a foggy lens, had a light leak in the camera, or shot through smoke or thick atmosphere. These are all similar scenarios where some fraction of the photons hitting the sensor have little to do with what's in front of the camera and are instead some single monochrome color. Gain instead represents the effect of if you had exposed for longer or if your lights were turned on brighter. Also both of these things can be used in the opposite direction too. If you have a lens flare you want to remove, a negative offset adjustment can handle that (it's very touchy though). If you overexposed, a gain between 0 and 1 can often be the best tool for the task in linear.
I've been watching your videos and learning a ton! But, as real estate video is my biggest source of income, I'd love to hear your take on how to maximize dynamic range, get sharp and contrasty looks, and beautiful saturated colors. It's a very different genre than anything focused on people and skin tones.
Cullen I really appreciate your efforts to the color comunity ( or "colour" uk people). I was part of your 1st workshop on tac. Did you make a limut on your 1st workshotp on tac?
I think that the confusion lies in the fact that linear is an adjective, and when used as an adjective has only descriptive meanings. So we need to listen to how it is used in context of what is being said. It's no different than saying a "curve" in a lot of things in cameras or Resolve. The confusing part is when it is used a title in gamma next to slog3, V-log etc. I have always used the word linear as one of two definitions - and I believe that if you use these two definitions - it's easier to know what someone is talking about when they are using the word "Linear". Just like your waveform of the greyscale ramp -the values are arranged in or extending along a straight or nearly straight line. But when used in defining stops or stages of data ( camera digital capture ) It's a progression of sequential steps from one stage to another in a single series of steps. On your greyscale ramp if you change the gamma to "Linear" you can see that it is in steps ( literally )
Well literally linear as an adjective should only refer to a line, not its shape. Straightlinear should be the appropriate technical term; from simple geometry seemingly it gets even more confusing when in math graphs "linear" may be referred to a "constant" property: a straightlinear progression in the graph from all the points mapped in the x and y axis which indicates a fixed relation of x and y no matter its values quantity. So it's really confusing as Gain Offset and Lift are not only keeping the straightlinearity of a gamma graph but also operating with mathematical constants, i definitively need a math class haha
@@IsaacRC I commented on how "I" interpret Linear when people refer to things in Resolve. When used as an adjective, it can be used for almost anything.
Specific Question for the community: Im working color managed in DWG. When I drop a ProRes4444 graphic clip with ALPHA in my timeline, I right click and select "bypass color management" so they are left alone a look like they did when they were handed to me. so they are bypassed form CM and placed ON TOP of everything and NOTHING is on the timeline level. However, they are not the same. On my vector scope they are much less saturated and slightly shifted. Weirdly, when I turn off all the clips below the GFX clip - they become normal. Full saturation and the correct color. Any ideas what is happening here? Thanks everyone and thanks Cullen for everything you're doing!
not sure if you figured it out already, but in case not: Davinci seems to have a problem with overlapping clips and colormanagment. Whenever you have a bypassed clip on top of a non bypassed clip it will take on the colormangement of the clip that got colormanaged. You actually need to give the prores a colorspace to fix it. For graphics you probably choose rec709 or srgb hope that helps. no idea why maybe somebody with more knowledge will help out here