[x265] Under VBV rc mode , increasing the vbv-maxrate or vbv-bufsize has no or very little effect on the final result

[Aarthi Priya Thirumalai]

Adding to Mario's points, changing the vbv-maxrate and vbv-bufsize , while
keeping the target bitrate the same - will have a heavier impact on the
moving average bitrate rather than the overall bitrate.
Increasing the vbv buffer, while keeping the rest constant roughly means
the buffer can handle a larger fluctuation in the per second bitrate.
Similary, decreasing the vbv buffer size will drive the encoder to a stick
to a constant bitrate for every second. At this point the vbv restrictions
will kick in to decrease the bitrate hikes in between,
and the quality for that portion of scene might go down.

So, I suggest you plot the distribution of bitrates for each of your test
case to see the impact .

Regards
Aarthi


On Tue, Oct 20, 2015 at 11:53 AM, Mario *LigH* Rohkrämer <contact at ligh.de>
wrote:

> I believe the reason is the relation between the three values. You can
> fill the buffer faster than the video target bitrate, so it needs only few
> seconds to fill a previously empty buffer, and even less if the buffer was
> already pre-filled. And then you compare it to a case where you can fill it
> even faster than before, so there is even less problem; the encoder is
> almost unrestricted by the VBV, only restricted by the target bitrate.
>
> You will certainly see a difference in cases where filling the buffer
> would take longer, and the video bitrate would exceed this fill rate for a
> longer duration if it was not limited by the fill rate: Then the encoder
> needs to reduce its target bitrate and spend less than the optimum for the
> desired quality.
>
> From my experience as DVD Video author, we had cases of close to optimum
> utilizations of videos just even rejected by the authoring tool, where the
> analysis of the scene in question revealed per GOP spikes of 13-14 Mbps
> (VBV max fill rate of DVD Video is 9.8 Mbps) which the VBV simulator of the
> encoder found acceptable, but the VBV simulator in the authoring tool could
> not accept (possibly due to an unfortunate burst of subtitles). Agerage
> bitrates tell little, their distribution over the VBV filling duration
> makes the difference.
>
> One of the most demanding scenes ever was the beginning of "Cruel
> Intentions" with a flight over a graveyard, filmed downwards: Several
> seconds of heavy motion right at the start. It revealed a few rare DVD
> player models which possibly had less than the specified 224 KB of decoding
> buffer built in.
>
>
>
> Am 20.10.2015, 04:59 Uhr, schrieb fangzhen <wintelx264 at 163.com>:
>
> Recently, I am testing the x265 codes  and mostly on VBV rc mode.\
>> But, I find an strange problem. For instance, I set up the following test
>> circumstance:
>> resolution: 1920x1080
>> bitrate=1000kbps, vbv-bufsize=3000kbit
>> when I set vbv-maxbitrate>1250kbps, I get almost the same result, like
>> the same psnr, the almost same bitrate.
>> I get the same result for vbv-bufsize where the vbv-maxbitrate is fixed,
>> like 3000kbps.
>> So I wonder why this situation exists? Can anybody give me some
>> explanation? Thanks so much in advance.
>>
>
>
> --
>
> 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/20151020/cd6d113c/attachment-0001.html>