[x265] Tiles support for x265
pleaselou at gmx.net
pleaselou at gmx.net
Wed Sep 14 10:54:23 CEST 2016
Ok, thanks for the information. Your are right that multiple slices per
frame are recommend
for most encoding applications. However, tiles can be used to realize tile
based streaming
(e.g. for VR/360), where each tile can be adapted in quality. For this it
would be also nice to
have control over the bitstream structure, to be able to have each
slice/tile in an own NALU.
2016-09-13 8:48 GMT+02:00 Mario *LigH* Rohkrämer <contact at ligh.de>:
>
> Jamaica posted a picture describing the different modes pretty well:
>
> http://forum.doom9.org/showthread.php?p=1780469#post1780469
>
> I am especially glad to have an imagination of WPP now. And even if Main
> profile does not support tiles in conjunction with WPP, higher profiles
> may. But so far for a long-term outlook...
>
>
>
> Am 12.09.2016, 22:24 Uhr, schrieb Tom Vaughan <
> tom.vaughan at multicorewareinc.com>:
>
> Sorry for the short reply. We're at the IBC Show, and it's keeping us on
>> the run from morning until late at night. I at least wanted to respond to
>> the question.
>>
>> Yes, we're working to implement full support for multiple slices per
>> frame,
>> and this will be followed by support for tiles. Things like rate control
>> have to be readjusted and tuned, and other algorithms may need to be
>> adjusted, so it may take a number of weeks to fully implement and
>> optimize.
>>
>> Multiple slices per frame is not recommend for most encoding applications.
>> As you cannot perform inter-prediction across slice boundaries (from frame
>> to frame), encoding efficiency is reduced. Multiple slices is mainly
>> beneficial for low-latency live encoding applications.
>>
>> Your description of slices and tiles below is basically accurate.
>>
>> Tom
>>
>> -----Original Message-----
>> From: x265-devel [mailto:x265-devel-bounces at videolan.org] On Behalf Of
>> Mario
>> *LigH* Rohkrämer
>> Sent: Monday, September 12, 2016 8:54 AM
>> To: Development for x265
>> Subject: Re: [x265] Tiles support for x265
>>
>> Shortest possible reply... :-P
>>
>>
>> I *guess* the recently introduced implementation of several slices is
>> already a step towards this even narrower approach.
>>
>> IIRC, slices already existed in previous video standards (like MPEG 1/2
>> Video) but were mainly meant to span across the whole width, or at least a
>> horizontal sequence of consecutove macroblocks in encoding order. Tiles
>> are
>> specifically designed to split the frame into a grid of rectangular parts,
>> like a "video wall" is made of a grid of monitors.
>>
>> Did I imagine correctly so far? ... It might be suitable to explain this
>> concept a bit in a kind of "outlook" post in a video forum, maybe even in
>> comparison with WPP (dis- vs. advantages).
>>
>>
>> Am 10.09.2016, 21:16 Uhr, schrieb Tom Vaughan
>> <tom.vaughan at multicorewareinc.com>:
>>
>> Yes.
>>>
>>>
>>>
>>> *From:* x265-devel [mailto:x265-devel-bounces at videolan.org] *On Behalf
>>> Of * pleaselou at gmx.net
>>> *Sent:* Saturday, September 10, 2016 2:08 PM
>>> *To:* x265-devel at videolan.org
>>> *Subject:* [x265] Tiles support for x265
>>>
>>>
>>>
>>> Hi!
>>>
>>>
>>>
>>> I just browsed the online docs
>>> (http://x265.readthedocs.io/en/default/)
>>> for
>>> information on how to encode a video using HEVC tiles and couldn't
>>> find anything. Are there any plans to implement this feature in the x265
>>> lib?
>>>
>>>
>>>
>>> I have to encode a video given a specific tiling pattern and doesn't
>>> want to use HM encoder cause it's pretty slow.
>>>
>>>
>>>
>>> Best, Mario
>>>
>>
>>
>>
>
> --
>
> Fun and success!
> Mario *LigH* Rohkrämer
> mailto:contact at ligh.de
>
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20160914/8d7cab57/attachment-0001.html>
More information about the x265-devel
mailing list