I cannot imagine how much work went into researching and condensing all of this. Thank you for doing this. Dope explanation. It’s surprisingly hard to find comprehensive information on this topic. You have the magic touch!
Hi. I love your videos, you put a lot of work on them and they are very concise and informative. However, there is a fundamental mistake in this one you probably might want to revisit. Gamma correction is a curve applied to "cancel" the curve from a gamma-encoded image. Our non-linear perception is the reason for the encoding, not the correction. We want to see the images in ratios as close to the ratios present in the capture image as possible. And that means that our eyes will have to receive light in the same ratios the camera "saw". If all digital images were 32-bit float images, gamma correction wouldn't exist, as cameras would capture and store linearly, and displays shouldn't have to apply any curve to them. But since we're still using 8-bit integer images a lot, computers need a way to store them in that precision. Linear won't cut it, because of our non-linear perception: our eyes are more sensitive to shadows, and 128 levels of brightness for the darker parts of the image won't be enough to represent a smooth gradient, you'll se a lot of banding. So, images have to be accommodated to allocate more bitdepth for the shadows, leaving less precision for lighter parts, the ones we have less sensitiity to. That's gamma encoding, and that's the bending gamma correction cancels. You have all the pieces there in your video, but the leading idea is unfortunately wrong
And to address the question: "why linear images look wrong when displayed directy" that most of the people reading this might have at this point, the answer is simple: Monitors are applying a gamma-correction automatically, but since the images not gamma-encoded (when you dump linear to the display) you end up looking at a non-linear image.
Very well done. Every other gamma curve explanation I've seen has neglected some aspect (capture tech, display tech, human eyesight) of the gamma curve topic, so definitely learned some new stuff here today.
Thank you so much! that explained everything. My understanding is 1. Human eye's light sensitivity is lower than that of the camera's. (ex. 50% light in camera is about 25% in human eye) 2. Gamma Correction, is a process that uses a mathematical model to match the digital screen light levels to that of the human eye's. 3. As the result, Gamma correction evenly distribute the light level across a digital screen which appears to be natural lighting to the human eye.
I almost never comment on videos but I couldn't not do so here. This is one of the best tutorial videos I've ever seen breaking down a complex, somewhat esoteric topic into an understandable format to folk who have little or no prior experience in the domain. You've earned yourself a subscriber, and if you have a link to send a few bucks to you or a charity of choice, please feel free to reply with a link. Thanks for this superb content.
I barely get overwhelmed! But this video is pure gold! Very well explained! And you my man, you have a bright future in front of you cause you can't explain things that properly unless you have both the knowledge and intelligence!
This is an extremely helpful, and remarkably well-constructed lesson. I confess to having been confused about this subject in the past, and now it is making sense to me. Looking forward to watching more of your stuff.
ERRATA: Adding ONE light, as in the example, does not DOUBLE the amount of light (or photons) hitting the paper, this is because of something called the inverse square law. In order to DOUBLE the amount of photons it would require adding THREE additional lights. Please correct me if I am wrong. Otherwise, thank you for a very well explained video on Gamma curves. Kudos for the 4K upload as well!
I'm not sure how the inverse square law applies in this case. In my demonstration, I'm assuming the two lights are perfectly pointed at the exact same spot, in which case there would be twice as many photons reflecting off of that point. Granted, the alignment wasn't actually perfect in my practical demonstration, but the point still holds. If you take two identical sources of light and point them at the same spot, there will be twice as many photons reflecting off of it as there would be if there was only a single light source. However, while the number of photons is doubled, our eyes don't *perceive* the spot as being twice as bright
You're clearly very knowledgeable on this subject and are great at explaining the technical aspects in an understandable way. Just learned a lot. Thank you!
Thanks for the video, Camon! I learned a lot. Can you also help me understand the mathematics here a bit more? Specifically why does running the luminance values through a power expression distribute more data in darker areas?
Needed to find out. Been playing alot of Halo & starting to think why some colours seem washed out & many shadows seem to dimmed when comparing other people's games. Turn down the gamma & boom, looks more natural, noticed some fog I forgot years ago. 😆