Menu
▾
▴
#293 SCALERS : 2XPM
2XPM, based on the Scale2X idea, analyzes any pixel
relation with its eight neighbouring pixels, to guess
some diagonal lines and rounded edges and apply the
obtained information in the four resulting pixels.
There are two versions of this scaler, HQ and LQ.
HQ is the "full-featured" version, its main purpose is
to enhance image quality, withouth worrying about CPU
usage. It uses different levels of alpha-blending to
enhance image quality when necessary.
LQ is the low-quality version, which detects most of
the lines detected in HQ, making it faster but the
resulting image will be more pixelated. Also,
alpha-blends are limited to 50% only, which is
optimized. LQ aims for speed, HQ aims for image quality.
The programmer of this scaler says that it's a
work-in-progress and unfortunately he hasn't developed
a 16 bit version of the scaler, I am sure you guys can
work around with it.
If you have any other inquiries, e-mail him at
pjmedina3@yahoo.com with the subject 2XPM.
Logged In: YES
user_id=12935
See <http://sourceforge.net/tracker/index.php?
func=detail&aid=1326624&group_id=37116&atid=418822>
However, we are extremely unlikely to add it at this time. So far, people on the
team perceived little to no gain by this scaler, and we already think we have to
many of them :-).
Logged In: YES
user_id=596410
Indeed, I agree there are already too many filters in
ScummVM. The thing is, some of them are honestly useless.
TV2x is crap in its current form. DotMatrix, who uses that
today? 2xSaI and SuperEagle are obsolete compared to
Super2xSaI or Scale2x. And HQ2x could easily be superseded
by 2xPM.
Problem is, I seem to remember Sev already closed a similar
request of mine. His argument was that 2xPM looked awful
compared to HQ2x. How is that possible, since in most cases
2xPM looks *exactly* as HQ2x? The main difference being that
HQ2x is totally useless on colourfully "busy" screens (think
MI2 here), while 2xPM is still able to smooth out the image
very efficiently.
I'm not doubting the quality of the patch. But the filter
works a treat in Kega, I'm not sure why it doesn't in ScummVM.
Logged In: YES
user_id=12935
For every scaler there is out there, there are people who
ask us to add it. For every scaler we add, there are people
out there who tell us we should remove it.
Guess what? People explicitly requested the tv2x and
dotmatrix scalers, for nostalgic reasons. They, for some
reason, enjoy playing with those. Personally, I don't
understand, just like you. But that doesn't change the fact
that some people think differently.
Some people love the hq2x scaler. Some think it's terrible
and ugly and unusable.
Some people love the 2xpm scaler. Some thing it's terrible
and ugly. Some think it's better than hq2x, some think it's
worse, some don't like either :-).
Personally, I've found 2xpm to be virtually the same as
hq2x, in some cases I found it to perform worse, in some
better, but in each only minimally so.
In the end, who's right? You argue that we should remove
something because *you* don't like it, and should add
something else because *you* like it. Shall I ignore what
everybody else says and put your oppinion above theirs?
Difficult choice to make...
Logged In: YES
user_id=596410
> Difficult choice to make...
I understand your position, but you hopefully agree that the
"first come, first served" basis can't be a viable road in
the long term :-)
Besides, I wasn't talking about what I like or not, I was
exposing plain facts (ok, I might be biased a bit ^^). The
scalers I mentioned are, IMHO, either obsolete (like
DotMatrix -- let's look at it, it's a fact, especially if
it's only used by a handful of people just for nostalgia
reasons) or redundant (2xSaI and SuperEagle have absolutely
no use since they're older generations of Super2xSaI -- at
best, they all look the same and they use the same CPU
resources).
As for 2xPM versus HQ2x, I exposed my arguments based on the
original 2xPM implementation in Kega. Here, my experience is
that 2xPM is a noticeable enhancement over HQ2x. But
ScummVM's mileage seems to vary indeed. And if people don't
like it, no problem, I rest my case. It's not as if it was a
priority for me (I'd much prefer to see Loom TGX supported,
even in 1x!).