View Single Post
Old 06-22-2019, 11:26 PM   #119
Staying Salty Staying Salty is offline
Special Member
 
Staying Salty's Avatar
 
May 2017
Earth v1.1, awaiting v2.0
Question What is Full Enhancement Layer (FEL) & Minimal Enhancement Layer (MEL) DV (FAQ)

Quote:
Originally Posted by Stacey Spears View Post
MEL is minimal enhancement layer. Means the video info is only in the base layer (10-bit 4:2:0) and the MEL only has metadata. FEL is full enhancement layer where the FEL layer has the extra picture information bringing the final output to 12-bit. Most DV content today is MEL. There is also the option where the FEL can be 4000 nits while the base layer is only 1000 nits. In our case, the base and FEL are both the max nit value.
Quote:
Originally Posted by Geoff D View Post
Dolby Vision encoding for disc is more complicated than DV encoding for streaming. I don't mean the actual DV master as that should be the same between both and is a separate process, I mean the physical encoding of that master into a disc or streaming format.

It's not that DV on disc is impossible to do, but it can take more time than a straight HDR10 encode, and more time means more money. If anything, the disc version of DV with FEL is the 'true' 12-bit decoded version while the streaming versions are all essentially MEL, there's no additional enhancement data whatsoever and they uprate the 10-bit stream to 12-bit purely via 'shaping' of the 10-bit version.
Quote:
Originally Posted by ShePearl View Post
So, you're saying 4K Blu-ray DV titles with MEL are 10-bit based with the 10-bit data padded into 12-bit container? (while FEL titles are the ones with 12-bit data splitted into two streams at the encoding stage and recombined at the decoding stage for full 12-bit reconstruction?)
Quote:
Originally Posted by Geoff D View Post
Yes.

[edit] though when you bear in mind that some content - especially older DIs - will have been finished at 10-bit anyway then one wonders where those extra DV bits are supposed to have come from, although a 12-bit HDR encode of a 10-bit HDR source should have less banding than a 10-bit encode of a 10-bit source. (Incidentally, there's a mastering mode in the Dolby Vision toolset which is specifically designed to take an HDR10 grade and Dolby Vision-ise it.)
Quote:
Originally Posted by Geoff D View Post
Aside from the improvements in tone mapping - which even the mighty ZD9 can benefit from in extreme cases, as we discussed - I don't think that there's a drastic difference between the actual HDR10 and DV grades at source level, particularly when the base layer is encoded at 4000 nits anyway. There is, however, an exception: Paramount's UHD of Annihilation which has an extremely black crushed base layer while the DV brings back the proper shadow detail.

However, I've still come to use DV more often than not because of how it has drastically improved the compression on some titles. Now whether the actual DV encoding process has knackered the HDR10 layer is up for debate (hi mike), but I find Saving Private Ryan to be unwatchable, literally unwatchable, in HDR10 while the DV improves it greatly. Ditto for several of StudioCanal's UHD titles, although their HDR10-only Rambos prove that they can encode HDR10 just as badly on a standalone disc as on a DV disc.

People aks how such voodoo is done and it's because of how the (full) enhancement layer rebuilds the image. Don't take my word for it, as Actual Professional™ Stacey Spears explained it in the HDR Discussion thread: the (full) enhancement layer is the result of comparing the final compressed HDR10 encode to the DV source master, so it does its best to replicate what it finds in there, be it differences in grading (as with the aforementioned black levels, or that the DV layer can contain >1000 nit information if the HDR10 layer has been capped at 1000 nits) or rescuing terribad compression.

These are things that HDR10+ is simply not capable of doing, though one could of course argue that if something was compressed properly in the first place then it wouldn't need DV to act as a firefighter in dousing them blocky-assed flames...
Edit:

https://www.avsforum.com/forum/465-h...l#post58228308



Quote:
Originally Posted by Geoff D View Post
Excellent find Salty! Not too sure there's a whole lot there that hasn't been enumerated by members of this forum, but it's interesting to see that 'pros and cons' table laid out like that. And the document reiterates that encoding DV for disc needs more work both before and after encoding, NOT that it's mission impossible, it just takes more time and more time means more money. It also means that they can't turn the discs around in the same time as what HDR10 'standalones' can be.

The P's & C's highlights just why Warners and Sony have rarely used FEL because they usually master their base encodes to 4000 nits and Dolby don't actually recommend using FEL with such base layers. One of the few Warners ones that uses FEL is 2001:ASO...which has a base layer of 1000 nits, not 4000. Interesting that the Matrices were also mastered with 1k nit base layers but didn't have an FEL.

Note that when it comes to DV content at 50/60p for disc then MEL is recommended, presumably because it'd be too much data to process efficiently. Seeing as Paramount have Ang Lee's Gemini Man and that it's bound to get a 4K60 UHD release with DV (becuz Par love dat DV) then it *should* carry an MEL version.
Quote:
Originally Posted by Geoff D View Post
Spot on. What some folks don't realise when they see those tiny MEL layers is that the DYNAMIC METADATA IS STILL THERE and is the same between both FEL and MEL versions. Metadata is something that takes up a literal fraction of space and isn't even part of a separately encoded layer on streaming DV, it's encoded into the least significant bit of the video image itself (the data dithered so as not to impinge on the video signal and encoded in triplicate for purposes of redundancy).

Yes, having a FEL does mean that the full 12-bit depth is present - improving banding as you say - and that it may well contain actual picture information that the 1000-nit HDR10 layer does not, but I've only seen that happen once across dozens of DV discs (and even then it was shadow information and not highlight detail!). DV FEL has also been a huge help on some discs that have very poorly compressed HDR10 layers but they shouldn't have been encoded that badly in the first place!

And besides, if the base layer has already been encoded to 4000 nit mastering display level then there's usually not a lot more for the DV layer to actually uncover. This is something that even Dolby has acknowledged, in the 2019 best practices guide they actually say that if you want to keep the base layer to 4000 nits then they RECOMMEND the usage of MEL, ditto if the content has only been mastered to 1000 nits at source level:
Quote:
Originally Posted by Geoff D View Post
It doesn't "cut that extra data". A true 12-bit output will give better gradations of colour (not more colour but finer shades thereof) even on a 10-bit panel because you're starting with a better source, it's basic oversampling theory.

The DV FEL layer contains informazione separate from the base layer which can affect the actual visual performance of the content, like improving compression and alleviating banding. And if there are >1000-nit highlights then they will be present in the tone mapped 900-nit DV output, unlike the base layer if it's been capped at 1000 nits.

Besides all of that, you're not thinking ahead: part of this is there to future-proof Dobly as an HDR format, so if/when we do get 12-bit displays that can do 4000 nits of peak brightness then several UHD discs will be sitting there smiling, saying "better to have it and not need it, than to need it and not have it".
Quote:
Originally Posted by MisterXDTV View Post
The bitrate drops the same way even with Dolby Vision. Just a little higher because of the additional FEL layer...
the base layer is the same even when you watch with DV
Quote:
Originally Posted by Geoff D View Post
^But you know how FEL works and what it's doing here, right? The bitrates still drop like a stone on a graph, yes, but in motion the FEL usually fills it in to the point where the compression becomes a minor annoyance (and what's my "minor annoyance" is utterly indistinguishable to most other folks) rather than a major disturbance, or is outright fixed even.
Quote:
Originally Posted by Streetlight
Can Dolby Vision MEL improve compression, or is it just FEL?
Quote:
Originally Posted by Geoff D View Post
I dunno about compression per se but in the Spears & Munsil 4K montage there’s a shot of a peacock’s feather which in HDR10 on my telly has some fine colour noise buzzing about in the “eye” of the feather, not the blocky chroma shite I usually reference but proper video noise. Playing the same bit with Doblys FEL and MEL eradicates the noise. Perhaps it has something to do with the internal processing of the DV engine performing its computations in ITP colour space? If so then there are still advantages to MEL other than the dynamic metadata.
Quote:
Originally Posted by PeterTHX View Post
MEL DV. HDR10+ is a squirt of metadata from an automated pass.

DV also has FEL which adds 12-bit capability and additional data.

Last edited by Staying Salty; 01-01-2022 at 02:11 AM.
  Reply With Quote
Thanks given by:
Angry Virginian (02-10-2022), cdth (10-30-2019), Nitroboy (06-23-2019)