Dolby Vision is poorly documented but I did find a long PDF explaining how it works.
It does provide significant value worthy of patent protection.
The main thing they did was develop a nonlinear color space designed so that each “bit” of information provides equal value. This way no bits are wasted, making compression more efficient and have fewer artefacts.
The color space is also set up so that the “lightness” channel is accurate and brightness can be rescaled without introducing color shifts.
They also came up with a way of encoding the HDR brightness range efficiently so that about 12 bits worth of data fits into 10 bits.
The format also allows a mode where there is an 8-bit SDR base stream with a 2-bit HDR extension stream. This allows the same file to be decoded as either HDR or SDR by devices with minimal overhead.
Last but not least they work with device manufacturers to make optimal mapping tables that squeeze the enormous HDR range into whatever the device can physically show. This is hard because it has to be done in real time to compensate for maximum brightness limits for different sized patches and to compensate for brightness falloff due to overheating. Early model HDR TVs had to have FPGAs in them to do this fast enough!
While Dolby did design the perceptual quantizer gamme (PQ) [1] that almost every HDR device today uses, they waived patenting it [2] when it was standardized in SMPTE 2084 [3]. Everything that is proprietary about Dolby Vision (everything except PQ gamma) is relatively mundane and just dynamic metadata.
People assume that some RGB matrix transform and a nonlinear PQ gamma is all Dolby Vision is.
In practice it's an entire ecosystem of certifications, professional calibration of panels, efficient encoding formats, etc...
To reproduce the end effect of Dolby Vision you'd have to have a team of people liaising with television manufacturers, and makers of production software like DaVinci Resolve.
It's not a trivial task that can be done through open source. It's real work that costs money.
We can all hope and wish for open standards, but it's a bit like trying to come up with an open architecture for a railway bridge. It'll still take real work to customise it to any particular valley, the local geography, and the specific requirements. Dolby Vision is similar. The mapping from the full signal range to each specific panel is a complicated thing that requires quite a bit of work to determine.
You obviously know more about this than I do, but when I watch a a well encoded Blu-ray that doesn't use Dolby Vision like Planet Earth 2, and I compare that to a dolby vision encoded disc, I fail to see a noticeable difference. And when I rip the RPU of a disc that uses dolby vision and look at some of the data [1], I again don't see enough to warrant this being a proprietary system, especially when we have HDR10+. So Dolby may have teams of calibrators and created certifications, at the end of the day I fail to see why this warrants being proprietary. Please let me know what Dolby Vision does better than HDR10+ or regular HDR, because at the end of the day I'm just a hobbyist and want to learn more about this space.
Dolby Vision is effectively 12 bit while using only 10 bits for encoding the actual signal. HDR10 is effectively... 10 bit. To achieve the same 12-bits of dynamic range they'd have to come up with a HDR12 format or something.
You can think of 8-bit SDR brightness as something like 0 nits to 255 nits of brightness. This is technically wrong because it's a nonlinear curve, but ignore that for a minute. Increasing this to 10 bits like with HDR10 gives you 0..1,023 nits with the same "steps". Going further to 12 bits lets you take this to 4,096 nits while continuing to preserve the same level of gradation.
The hiccup with this is that some displays have 600 nits of peak brightness and some have 2,500 nits. There are rare displays that can go to 4,000, and prototype displays that go to 10K nits.
HDR10 only goes to 1,000 nits.
Dolby Vision goes to 4,000 nits.
Does this matter now on some cheap LCD TV that only goes to 600 nits? Probably not.
Does it matter on an OLED panel that only goes to 1,000 nits? Maybe, because Dolby Vision has more true-to-life mapping from the maximum range to the display panel range.
Will it matter more with next-generation panels capable of 3,000+ nits? Almost certainly.
Then again, HDR10+ has dynamic metadata, which compensates a lot for its lower bit depth. Additionally, most smart televisions smooth the "steps" in smooth areas, largely eliminating the artefacts caused by HDR10.
At the end of the day, they're both significantly better than SDR, but Dolby Vision is a touch better, especially on high-end panels.
Based on my own research, I don't believe this information is correct. HDR10 video is encoded according to Rec 2100 [1], which states that the video is 10 bits and goes up to 10,000 nits. I know this to be true as I have written my own programs to encode HDR10 video. You are correct however that most HDR10 content is only encoded with a mastering luminance of 1000 nits [2]. According to this [3] forum post, all DV content (besides profile 9) is encoded using 10-bit HEVC stream. When people talk about 12-bit Dolby Vision, they are referring to Profile 7, which contains a full enhancement layer stored at a 1/4th the resolution, which contains the residual between the source and the Base Layer, allowing a player to decode the data to 12 bits. It appears [4], however, that most DV content only uses the minimal enhancement layer, which is only the dynamic metadata.
I was oversimplifying on purpose. The standards are complex and have multiple modes. These interact in weird and wonderful ways with hardware capabilities.
Which particular content have you noticed a big difference on?
I have an (older) LG OLED, and haven't seen anything in DV that I didn't think would be just as good in HDR10, although I haven't compared the same content in both DV and HDR10 directly.
I was under the impression that there were no Dolby Vision reference monitors, and that different players, like Sony and Oppo, output Dolby Vision encoded 4K Blu-rays differently, and thus there is no ’correct’ interpretation of the metadata.
This is opposed to HDR which has reference monitors and should look exactly the same on different, calibrated systems.
It does provide significant value worthy of patent protection.
The main thing they did was develop a nonlinear color space designed so that each “bit” of information provides equal value. This way no bits are wasted, making compression more efficient and have fewer artefacts.
The color space is also set up so that the “lightness” channel is accurate and brightness can be rescaled without introducing color shifts.
They also came up with a way of encoding the HDR brightness range efficiently so that about 12 bits worth of data fits into 10 bits.
The format also allows a mode where there is an 8-bit SDR base stream with a 2-bit HDR extension stream. This allows the same file to be decoded as either HDR or SDR by devices with minimal overhead.
Last but not least they work with device manufacturers to make optimal mapping tables that squeeze the enormous HDR range into whatever the device can physically show. This is hard because it has to be done in real time to compensate for maximum brightness limits for different sized patches and to compensate for brightness falloff due to overheating. Early model HDR TVs had to have FPGAs in them to do this fast enough!