Official Off-Topic Thread 2014

[Dusty]

Dusty:
No need to imagine! Just look at the Super Mario World clip on their website -- In the 'Video Comparisons' section, click on 'Our Result'. The clip is limited in area, but certainly has enough to get a decent idea.

Problem is Mario is one of those cases that doesn't have a lot of pixel busyness going on, and thus benefits pretty well. I'd like to see another game like Secret of Mana, Final Fantasy 3 or some such that has dithering and noisy textures and such.

[Carnivac]

Dunno.  The Mario example still looked bloody terrible to me.  None of these things ever look even a fraction as good as the original graphics.  Makes me wonder why bother playing old games if you're just gonna turn their well crafted sprites into such crud...

[Ai]

Fancy upscaling is more about necromancy than preserving the aesthetic, though. Considering graphics bad if there are 'visible pixels' is definitely a thing. Not a thing we entertain at Pixelation, but a consumer thing. Eliminate the pixelization ~= eliminate the badness, without drawing new assets! 

[Mathias]

Great work, Carnifax.

 

Maybe he thinks your name is a reference to the metal band: Carnifex?

Or the Latin word for "executioner"?

Or I could just be overexamining a simple error :p

Haha . . .

Or maybe I was critically sleep-deprived, as usual. Good work, Carvinack!

[Helm]

Carvinack!

It's really not that difficult.

[Ymedron]

Carving snark?

[Dusty]

Dunno.  The Mario example still looked bloody terrible to me.  None of these things ever look even a fraction as good as the original graphics.  Makes me wonder why bother playing old games if you're just gonna turn their well crafted sprites into such crud...

Sadly everyone doesn't appreciate pixels as much as we do

[YellowLime]

Carvinack!

It's really not that difficult.

Carving snark?

(come on people you know the second time was on purpose)


This pixel-vectorizing business is pretty interesting to me. Not because I dream of a dystopian future free of pixels, but because I want to know the thought process in trying to make it look as good as possible (or the least awful ) Since, you know, code is art-blind.

It also makes me wonder how close can these methods get to creating nice graphics, and what would be the thinking/planning behind it.

Why don't any of the algorithms work with user input? I guess that since filters are supposed to be used at runtime, they don't bother to ask a user's opinion (I mean, it's like they're used on the "consumer" side)

But if someone were to partake in the unholy activity of purposefully "porting" pixel graphics into vector counterparts, any algorithm would benefit from the user's aesthetic criteria. (so, a filter with "options" for the "developer" side.)
(Though not so much "options" but instead "mapping qualities to specific groups of pixels")

In fact, I find it weird that there wouldn't be any filters like that! Maybe you guys know of any?

[Dusty]

I think the problem will always be that algorithms aren't smart enough to interpret pixels. A huge part of pixel art is that one pixel can change what we see as a whole, and algorithms will never be able to interpret and properly scale up what that pixel is supposed to be. I think the first step would be to start taking into consideration the layouts we use as pixel artists to make shapes.

I think out of all those examples(here) hq4X is the best. It has its flaws, but it does a pretty damn good job most of the time in interpreting what shapes are being created with the pixels, and to top it off has some nice AA. If only that could be refined, as it seems to "skip" many pixels and just square them off.



But at the end of the day, it can't look at the overall set of pixels and interpret what the artist intended, and fails as well. Again, the more complicated the pixels, the more it seems to fail. And I don't think we'll really ever overcome that.

[Ai]

xBR seems to achieve better results than HQ4x.
Its author, Hylian, has posted here about it (but has improved the algorithm since then.) . In the link I gave, he outlines the challenges of recognizing finer increments in angles -- in particular, that the 'output area' may go outside the area that you are reading to detect the angle (so the algorithm will want to overwrite its own results in places). Despite this he's working on a 'level 3' filter (level 1 == 45 degree increments, 2 == 22.5 inc, 3 = 11.25 inc). If you poke him he might have something more to say on this current topic here.

BTW: your link is strange --

Code: [Select]

http://"https//research.microsoft.com/en-us/um/people/kopf/pixelart/supplementary/multi_comparison.html%22 ?