View Single Post
Old 16th September 2019, 15:24   #156  |  Link
FranceBB
Broadcast Encoder
 
FranceBB's Avatar
 
Join Date: Nov 2013
Location: Royal Borough of Kensington & Chelsea, UK
Posts: 2,883
Quote:
An Adaptive Multiple Transform (AMT)
method has been designed such that depending on the selected mode (intra or inter)
Got it.

Quote:
In addition to AMT, a Mode-Dependent Non-Separable Secondary Transform (MDNSST) is applied
between the core transform and the quantization, with the motivation to reduce remaining dependencies after the
separable core transforms which only address horizontal and vertical signal dependencies.
Ok, it makes sense.

Quote:
Finally a Signal
Dependent Transform (SDT) is competed to the AMT output. The SDT approximates the optimal Karhunen-Loéve
transform (KLT), which is a signal dependent transform, by estimating current signal to code (transform block) with
similar signals (i.e. reference patch) available at the decoder (already coded).
Very interesting 'cause this way a lot of computational power is actually saved by not using the KLT directly which is far too demanding in terms of computational cost, but if this approximation gets similar results, I totally understand why they went for this path.

Quote:
AV1 also supports multiple transforms: DCT, Asymmetric DST (ADST), flipped ADST, and Identity (IDTX).
Oh, I see... I gotta say that I'm not very familiar with AV1 'cause it's not very widespread in broadcast and nobody ever sent us a master file in AV1, but it's good to know.

Quote:
The
identity transform is equivalent to the transform skip mode of HEVC and JEM.
Yeah, of course, it's just like when you multiply a matrix of linear transformation by the identity matrix (the matrix with all one on the main diagonal and zero everywhere else): you get back the matrix you started from. After all the identity it's there (in a space) for a reason: to get back the input as output

Quote:
The vertical and the horizontal
transforms can be selected independently from the set of four transforms, resulting in 16 transform combinations.
I see, that's good.

Quote:
the square transforms in HEVC are extended to non-square transforms for rectangular blocks resulted from binary and ternary tree splits.
I see! That's gonna increase the computational cost, but it's actually very good that they have been modified, so if they did it, it means that the result it's worth the cost.

Quote:
CABAC coder from AVC, it has been enhanced and is now even faster.
Oh, that's great.


Alright, thank you for clarifying my doubts. I'm an young engineer and I have always been interested in applied mathematic. VVC has a lot of potential indeed and some of those things are really reassuring me about VVC and the role that it's gonna play in the future.
Speaking of future, Tokyo 2020 Olympics are gonna be in 8K and we're talking about summer 2020. Do you think that the encoder will be at a much more developed state at that point and it's actually gonna be used to encode and distribute live feeds or do you think that it's still way too early and everything it's gonna be done in HEVC as rumours suggest?

Last edited by FranceBB; 16th September 2019 at 15:28.
FranceBB is online now   Reply With Quote