[x265] Ghosting/artifacts even with low CRF

Mario Rohkrämer contact at ligh.de
Fri Jul 15 21:16:19 CEST 2016


Just a nit: "--crf 0" is not meant to be "lossless", the quantization  
might still vary; "--qp 0" is certainly lossless, as documented in "x264  
--fullhelp". But you should consider your file "quasi-lossless", with no  
noticeable difference.


Am 15.07.2016, 20:25 Uhr, schrieb Rômulo Silva <rsilva2k5 at gmail.com>:

> Hello.
>
>
> There you go:
> http://www.mediafire.com/download/crmj9d7ylsg32dm/x264_crf0_hardsub.mkv
>
> Encoded with x264, revision 2692, 10bit, preset veryslow, CRF 0.
>
>
> Cheers,
>
> R.S.
>
> On Fri, Jul 15, 2016 at 8:56 AM, Deepthi Nandakumar <
> deepthi at multicorewareinc.com> wrote:
>
>> Hi,
>>
>> Sorry I couldnt get to this before v2.0. Can you share a file with the
>> subtitles burned into the x264-crf0 lossless video?
>>
>> Thanks,
>> Deepthi
>>
>> On Sun, May 15, 2016 at 3:49 AM, Rômulo Silva <rsilva2k5 at gmail.com>  
>> wrote:
>>
>>> Some more info on the issue.
>>>
>>>
>>> I updated my x265 encoder to 1.9+167 and tested the same sample with  
>>> the
>>> same parameters (CRF 21, 10-Bit, preset slower) with  
>>> --no-recursion-skip as
>>> the only additional parameter.
>>>
>>>
>>> x265 hardsub sample [--no-recursion-skip]:
>>> http://www.mediafire.com/download/lx98hc6n5b40db2/hardsub+test+norecursionskip.mkv
>>>
>>>
>>> The result is much better compared to not using the --no-recursion-skip
>>> parameter, but there's still some ghosting on both books in the scene
>>> around the letters. File size is around 10% bigger, and encoding time  
>>> was
>>> more or less 50% slower just like pointed out here:
>>> http://forum.doom9.org/showthread.php?p=1767551#post1767551
>>>
>>>
>>> Thanks again,
>>>
>>> R.S.
>>>
>>> On Mon, May 9, 2016 at 5:59 PM, Rômulo Silva <rsilva2k5 at gmail.com>  
>>> wrote:
>>>
>>>> This is related to issue #214:
>>>> https://bitbucket.org/multicoreware/x265/issues/214/ghosting-artefacts-even-with-low-crf-when
>>>>
>>>>
>>>> Turns out that it's easy to reproduce the problem with hardsubs, and
>>>> I've included both x264 and x265 samples below to illustrate this.
>>>>
>>>>
>>>> Lossless x264 sample:
>>>> http://www.mediafire.com/download/ftwysmcemw7211a/x264+crf0.mkv
>>>> Sample subtitle:
>>>> http://www.mediafire.com/download/jttvyk2k0va9c3g/lorem+ipsum.ass
>>>> x264 hardsub sample:
>>>> http://www.mediafire.com/download/nzp7pzz5y8h7yjy/hardsub_test_x264.mkv
>>>> x265 hardsub sample:
>>>> http://www.mediafire.com/download/mz51v2a8d99pa5t/hardsub+test.mkv
>>>>
>>>>
>>>> Both hardsub files were encoded with CRF 21, 10-Bit and slower preset.
>>>> X265 encoder version was 1.9+150-00ea3784bd36.
>>>>
>>>> You can clearly see the artifacts between the two lines on top of the
>>>> left book. Some ornaments get distorted around the letters, and  
>>>> straight
>>>> edges don't move according to the scene.
>>>>
>>>> A few suggested workarounds, like --no-cutree, make the ghosting less
>>>> apparent, but it still doesn't solve the issue completely. I've  
>>>> included
>>>> below a sample of the same x265 encode used above (CRF 21, 10-Bit,  
>>>> slower)
>>>> with --no-cutree as the only additional parameter.
>>>>
>>>>
>>>> x265 hardsub sample [--no-cutree]:
>>>> http://www.mediafire.com/download/abvwr1jbp47uw29/hardsub_test_nocutree.mkv
>>>>
>>>>
>>>> The resulting file has over twice the size, and even though the
>>>> artifacts on ornaments are mostly gone, some straight edges in  
>>>> between the
>>>> lines are still not moving according to the scene.
>>>>
>>>> Has any progress been made on this issue? I'm willing to help by
>>>> providing more samples and test cases, if needed.
>>>>
>>>>
>>>> Thanks for your attention,
>>>>
>>>> R.S.
>>>>
>>>
>>>
>>> _______________________________________________
>>> x265-devel mailing list
>>> x265-devel at videolan.org
>>> https://mailman.videolan.org/listinfo/x265-devel
>>>
>>>
>>
>>
>> --
>> Deepthi Nandakumar
>> Engineering Manager, x265
>> Multicoreware, Inc
>>
>> _______________________________________________
>> x265-devel mailing list
>> x265-devel at videolan.org
>> https://mailman.videolan.org/listinfo/x265-devel
>>
>>


-- 

Fun and success!
Mario *LigH* Rohkrämer
mailto:contact at ligh.de


More information about the x265-devel mailing list