[vlc-devel] In deinterlacing this week: odds and ends

Juha Jeronen juha.jeronen at jyu.fi
Fri Apr 8 14:11:41 CEST 2011


On 04/08/2011 10:44 AM, Jean-Baptiste Kempf wrote:
>> 3) After writing the summary table in the doc draft, including X (2x)
>> (which I plan to suggest for merging later), I noticed that now Linear
>> is the odd one out, as it's missing a 1x mode...
> Does it make sense in 1x mode?

In my opinion, as much as Discard does :)

(Although there is a subtle difference, Discard can be thought of as a
1x mode for Bob, keeping only the top field.)

>> However, that wouldn't work much better than Discard, so probably no one
>> would want to use it, unlike X (2x) which can be useful. Maybe I'll try
>> to resist the temptation to make everything symmetric ;)
> Find a nice asymmetric function and study it... Breathe... :D


I was thinking that if Linear had a 1x mode (and X a 2x mode), there
would be a logical progression of interpolator/doubler pairs with
different interpolation methods: none (Discard + Bob), linear (Linear),
edge-oriented (X) and spatial/temporal (Yadif).

But this would still leave Phosphor and IVTC as the odd ones out... 1x
for Phosphor doesn't even make sense, and for IVTC, I've been thinking
about this some more and 2x doesn't make sense in the context of the
current algorithm. (It would for Phosphate/IVTC2x, but there's probably
no point in developing that further.)

>> (Myself, I was puzzled by the menu entry "Yadif (2x)" back when I hadn't
>> read the source code.)
> Indeed.
> And X (2x) is not better.

Mm. That's just a working label to go with Yadif (2x) :)

2X? X at 2x? X<sup>2</sup>?


Y? :P


More information about the vlc-devel mailing list