[x264-devel] [Tom Jacobs <T.R.Jacobs at lboro.ac.uk>] Re: Re: tiles

[System administration]

 The deleted attachment is at:
    <http://www.videolan.org/~admin/20051003-videolan/br-slice.jpg>

----- Forwarded message from Tom Jacobs <T.R.Jacobs at lboro.ac.uk> -----

From: Tom Jacobs <T.R.Jacobs at lboro.ac.uk>
Date: Sat, 01 Oct 2005 11:34:06 +0100
To: x264-devel at videolan.org
Subject: Re: [x264-devel] Re: tiles
User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317)
X-Spam-Status: No, score=-3.4 required=5.0 tests=IN_REP_TO,RCVD_IN_ORBS 
	autolearn=failed version=3.0.3

here is a graph from my work on slices/threading which shows the bitrate 
affect of encoding using 4 slices on a QCIF, CIF and 1088p image. i can 
find the graph right now but for HD images up to 16 slices introduce 
bitrate increases of less than 10% 36db and 5% at 44db.

Tom



Loren Merritt wrote:

>On Fri, 30 Sep 2005, reamer-s at Safe-mail.net wrote:
>
>>On Fri, 30 Sep 2005, Loren Merrit wrote:
>>
>>>Just note that slice-groups (needed for any shape other than raster 
>>>scan)
>>>are allowed only in baseline and extended profiles.
>>
>>
>>My application is for encoding HD up to 1080i.  What is the impact of 
>>these profile restrictions?
>
>
>It means you can't use CABAC or 8x8 transform, which add quite a lot 
>of compression (together 15% bitrate or so).
>
>>>And no matter what, you'll still lose some quality because inter blocks
>>>can't predict MVs from across the slice border.
>>
>>
>>If the tiles each cover a relatively large portion of the frame does 
>>that help?
>
>
>If the tiles are large (especially at 1080i), then the bitrate lost to 
>border effects is comparatively small.
>
>--Loren Merritt
>



----- End forwarded message -----

-- 
System administration <admin at via.ecp.fr>
VIA, Ecole Centrale Paris, France

-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html