[FFmpeg-user] Benefits from Chroma Upsampling v4l2 8bit 422 to 10bit 422 FFV1 backup?

[Erik Dobberkau]

> "For video tapes, we always recommend capturing lossless, 10-bit, 4:2:2 or
> better"
> https://www.archivalworks.com/blog/best-practices-for-digiti
> zing-and-preserving-analog-media
>
> If you're trying make high-quality copies of analogue video, you really
>> need to treat the analogue signal before digitizing it - at a minimum there
>> ought to be a time-base correcter (TBC) in there;
>>
>
> Thanks for your feedback.
> And yes, I agree. I didn't tell that I actually have a legacy pretty good
> TBC (built in a Videonics MX-1 videomixer) which I also used before A/D
> converting and capturing DV25 on a Datavideo DN-300 DV & HDV HDD recorder.
> The MX-1 PAL version, which also had a usable noise filters, was specified
> as "broadcast quality" with 17,72 MHz sampling, 8-bit 4:2:2 (not 10-bit as
> its successor MXPro).
>
> So, I plan to apply the MX-1/TBC between the S-video player and the A/D
> converter the HDMI based chain above.
>
> My suspection is that the visual lesser color and brightness from DV25
> playback, mainly is due to the format's limited 8-bit 4:2:0 sampling.
>
>
> better would be a waveform monitor and proc amp* so you can adjust things.
>>
>> *for instance, if the white digitizes to 180 instead of 235 (or 254/5),
>> stretching that in ffmpeg loses intensity resolution as it has to
>> interpolate values.
>>
>
> This is beyond my knowledge.
> What I also had available, but possibly is non-working now, is a legacy
> video-color corrector (Video Tech Designs VCC 3010).
>
> (And not to mention, I also have a legacy Blackmagic A/D-SDI converter +
> Hyperdeck Shuttle2 SSD recorder, which managed 10-bit ProRes HQ 422 for
> both S-video and HDV.
> But last time I tried, these devices were unhappily non-usable due to EOS
> BM firmware)
>
>
> So therefore I am still curious in suggestions regarding the requested
> "10-bit Chroma upsampling" using ffmpeg and possible color noise filters(?)
> to benefit Analog (and also HDV) backup via HDMI?
>
>
+1 for Carl's comments.
Your results largely depend on how well you can 'massage' the signal in the
analogue domain, and then on the capabilities of your A/D converter,
especially its quantizer. If you can only capture at 8bpc, there is no
point in re-sampling the original capture file to 10bit, you're not storing
better information. If your capturing device gives you a 8bpc 4:2:2 output,
that's what I think you store store as your uncompressed/lossless master
file to work from later. Everything done afterwards ... well, YMMV. You can
still re-sample from 8bpc to 10bpc and have no visible or measurable
difference compared to a 10bpc capture from an 8bpc converter.
In some cases (even segments from the same tape source) you may achieve
satisfactory results with one setting/combination, others may require a ton
of adjustments and still look unsatisfactory.
I guess this is not the answer you would have loved to hear... but
(professional) archival and restoration is rarely a set-and-forget effort.