Quote:
Originally Posted by haertig
Of course it's not the same picture. One was taken with an SDR camera, and the other was taken with an HDR camera standing next to the first camera.
|
Ah, I assumed this was supposed to be a video frame taken from a UHD Blu-ray and its HD counterpart, and thus it should have been possible to grab the same frame number both times. My mistake.
Quote:
Originally Posted by haertig
and finally displayed on a device that is capable of handling the HDR info.
|
This is a wrinkle I was afraid of. Since I'm looking at the image on a TN LCD, my understanding is it's impossible for me to see the true benefits of HDR, anyway.
Quote:
Originally Posted by haertig
I think it is somewhat akin to taking multiple captures of an image, each using different exposure settings, and then combining those images taking the best from each and coming up with a composite.
I believe the techniques to come up with an HDR output in video are implemented differently.
|
Yes, HDR has nothing to do with creating a composite image. It has to do with how the image is stored. We usually think of digital images in terms of R G B values, but storing full RGB information requires way too much space for video. Instead, the video is transformed into chroma (colour) and luma (brightness) components. Since the human eye is much more sensitive to brightness than colour, the colour component is usually stored at quarter resolution, e.g. the luma is 1920x1080, but the chroma resolution is only 960x540. This is called chroma subsampling, of which 4:2:0 is the most common type.
The luma (Y) and chroma (UV) are commonly arranged in a planar format (meaning Y and UV are stored separately, rather than interleaved) called YV12. This is much more efficient than RGB, and the quality degradation is hardly noticeable, leading to a big compression efficiency increase, but there's a problem: how do you convert back to RGB, so you can actually display the image on your TV?
There are multiple standards for this, and this is where SDR vs HDR comes in. Regular Blu-ray uses
Rec.709 to define the rules of the conversion, whereas UHD Blu-ray uses
Rec.2100 for things like Dolby Vision. The main thing to note on each Wikipedia page is the spectrum area covered by the black triangle: Rec.2100 (and Rec.2020, which Rec.2100 was extended from) cover a much higher percentage of all possible visible colours than Rec.709 does by specifying different conversion formulas for how to move between YUV and RGB. Rec.2020 covers the same wide gamut as Rec.2100, but it doesn't support HDR.
As long as you record the original footage in HDR, you can convert to SDR by doing a YUV -> RGB conversion using Rec.2100, and then converting RGB -> YUV using Rec.709 (or Rec.601, if you want SD resolution).
The move from 8 bits per pixel to 10-12 bits per pixel is also somewhat relevant, but this is mostly useful from a compression perspective: the banding artifacts caused by using 8 bpp reduce compression so badly that using 10 bpp (4x more precision) actually improves compression, despite the need to represent more information.
Hopefully I haven't mucked this up too badly. Video encoding is a complex subject, and I'm only a hobbyist, not an expert.
