mike admin wrote: ↑Fri Oct 02, 2020 1:26 pm
I wonder if it is actually a profile 8 stream at all.
I hope Spears and his friend Munsil don't use my tool to create their BDs
.
I therefore assume that the stream was obtained with certified software.
According to this logic it must necessarily be a MEL stream profile 7.
mike admin wrote: ↑Fri Oct 02, 2020 1:26 pm
Interesting. What happens if we just discard the EL layer in this case? Do you think it is safe?
What Happens? A Dolby Engineer Dies
!
Seriously, this is my idea:
If you decide to mark the stream as profile 7, the hardware decoder will process everything through two distinct Threads (one for EL and the other for BL) sharing, simultaneously, the information contained in the same RPU. In the case of BL, the mapping parameters (for the color component cmp) are used to predict the final HDR stream. In the case of EL, the inverse quantization parameters are used to process the EL signal. Clearly, in the case of a FEL stream, a decoding is needed otherwise, in the case of a MEL, hardcoded informations are used (decoding is therefore not necessary). Finally, the BL-derived HDR stream is added to the inverse quantized signal of EL to recreate the original HDR stream.
For the operation to be "thread-safe" it is necessary that the two streams (EL and BL) are perfectly aligned. The RPU they share does not contain any spatial references and does not contain any IDs. This is why it is important that the NALUs of EL (even if practically empty in the case of MEL) are maintained. The "context" of RPU is deduced from the "surrounding environment" common to BL and EL.
So, in this case, I don't think it's safe to remove EL.
I hope I’ve made myself clear.
The other option is to mark the stream as profile 8: you have to delete EL and modify the RPUs (e.g. el_spatial_resampling_filter_flag should be set to false).
We will find out in the future what other parameters will need to be changed!