Gp2x Has The Biggest Screen Size Among Handhelds


junker posted on Oct 23 2005 at 07:22 PM said:
hey guys, just quit this discussion. it went no where last time!
No. DaveC FINALLY has to understand how it works. :angry:
 
Last edited by a moderator:
Eeek; really,something like the gp2x and GP32 BLU is almost too crisp; original monitors distorted and blurred, especially if your'e talking about old television sets like C64's and such were designed for. More to point, arcade monitors are almost required to blur.. they _expected_ and _depended_ upon the blurring effect :)

Why must people make platform arguments? (he asks, baiting someone to make an sT-vs-Amiga or vi-versus-Emacs dig :)

jeff
 
I actually trhink DaveC is one of the few who can 'win' (as winnable as an _opinion_ based argument can be) -- DaveC doesn't somuch care about the specs ot details.. he cares about it looking 'right'; any form of distortion (biliean is in fact distorted, though .. like a jpeg.) Right or wrong isn't relevent..he prefers it this way (be it decided or inherent preference, thats his business.) (And he likely enjoys jpegs at the same time, since we're not talking general image representation, but specifically game console representation.)

As an arcade purist, I've always said it can't _Really_ be done without true machines; you _cannot_ emulate vector games without a vector monitor, since no other kind of monitor can do near-infinite resolution with brightness control (real brightness, that burns out your eyeballs :); and really, unless oyu're rendering on a 25" blurry TV with wood panelling, it won't look _or sound_ right.

(Remember, back in the day, they actually assumed you had shitty speakers and wood enclosures etc. I know several machines designed to fit in wood or plastic TVs with crappy sound, to get what they achieved. Using emulation means it comes out too-crisp,and too-pure.)

But we don't all have room for a dozen consoles or arcade machines :)

So _everything_ is an approximation, and ad such you decide the degree to which you accept it. If sub-pixel-scaling is okay for you, good; if not, fine; all good. Just find a platform that fits your needs as close as possible if you like :)

jeff
 
I'm still scarred from the ST-Amiga war... I have a A500 mouse lodged near my neo-cortex. Removing it would risk permanent injury, even death. So I just wear a big hat.
 
OrR posted on Oct 23 2005 at 03:51 PM said:
No, you fucking idiot, there is absolutely NO distortion. :angry: Sorry about my language but I'm REALLY loosing my patience. Bilinear stretching like on the Zodiac is NOT pixel doubling + blur. That is complete bullshit. There is no distortion, every pixel counts as 1.5 pixels.
See, every pixel from the original counts the same. The edges are blurred together.
Get it now? :angry:


YOU are the fucking idiot. You DOUBLED the pixel in your example. 1.5 is NOT *double* it is 1 and a half times. YOU CAN'T LIGHT HALF A PIXEL ON AN LCD ! You example is shit because you don't show the grid representing the fixed pixel grid of an LCD you just resize a rectangle anywhere you want. It doesn't work that way.

I think we are miscommunicating here ;)

pixel.png
 
Last edited by a moderator:
DaveC posted on Oct 23 2005 at 11:01 AM said:
OrR posted on Oct 23 2005 at 03:51 PM said:
No, you fucking idiot, there is absolutely NO distortion. :angry: Sorry about my language but I'm REALLY loosing my patience. Bilinear stretching like on the Zodiac is NOT pixel doubling + blur. That is complete bullshit. There is no distortion, every pixel counts as 1.5 pixels.
See, every pixel from the original counts the same. The edges are blurred together.
Get it now? :angry:


YOU are the fucking idiot. You DOUBLED the pixel. 1.5 is NOT double it is 1 and a half times. YOU CAN'T LIGHT HALF A PIXEL ON AN LCD MORON!!!! God you are so dumb. You example is shit because you don't show the grid representing the fixed pixel grid of an LCD you just resize a rectangle anywhere you want. It doesn't work that way.

You are the idiot. He is speaking of interpolation. It means to blend the adjacent pixels together when necessary to produce the 1.5 pixel effect.
Heres a blown-up example.
pixelthingy3ex.png

On the LCD screen, the 2 pixels are actually 3 (higher res) so this can work.
 
Last edited by a moderator:
Blah posted on Oct 23 2005 at 06:08 PM said:
DaveC posted on Oct 23 2005 at 11:01 AM said:
OrR posted on Oct 23 2005 at 03:51 PM said:
No, you fucking idiot, there is absolutely NO distortion. :angry: Sorry about my language but I'm REALLY loosing my patience. Bilinear stretching like on the Zodiac is NOT pixel doubling + blur. That is complete bullshit. There is no distortion, every pixel counts as 1.5 pixels.
See, every pixel from the original counts the same. The edges are blurred together.
Get it now? :angry:


YOU are the fucking idiot. You DOUBLED the pixel. 1.5 is NOT double it is 1 and a half times. YOU CAN'T LIGHT HALF A PIXEL ON AN LCD MORON!!!! God you are so dumb. You example is shit because you don't show the grid representing the fixed pixel grid of an LCD you just resize a rectangle anywhere you want. It doesn't work that way.

You are the idiot. He is speaking of interpolation. It means to blend the adjacent pixels together when necessary to produce the 1.5 pixel effect.
Heres a blown-up example.
pixelthingy3ex.png

On the LCD screen, the 2 pixels are actually 3 (higher res) so this can work.

That causes color distortion he says it can do it with no distortion. He says NO distortion That is false. A color is introduced that wasn't even in the original image.

Yeah you can use interpolation but it makes the edges look fuzzy. That is what I said in the beginning is that when you stretch at an odd factor you get distortion. Anything added or taken away is considered distortion.

Whether you like it or hate it is subjective.
 
Last edited by a moderator:
DaveC posted on Oct 23 2005 at 08:01 PM said:
OrR posted on Oct 23 2005 at 03:51 PM said:
No, you fucking idiot, there is absolutely NO distortion. :angry: Sorry about my language but I'm REALLY loosing my patience. Bilinear stretching like on the Zodiac is NOT pixel doubling + blur. That is complete bullshit. There is no distortion, every pixel counts as 1.5 pixels.
See, every pixel from the original counts the same. The edges are blurred together.
Get it now? :angry:


YOU are the fucking idiot. You DOUBLED the pixel in your example. 1.5 is NOT *double* it is 1 and a half times. YOU CAN'T LIGHT HALF A PIXEL ON AN LCD MORON!!!! God you are so dumb. You example is shit because you don't show the grid representing the fixed pixel grid of an LCD you just resize a rectangle anywhere you want. It doesn't work that way.

YOU need to understand how a FIXED PIXEL display works, obviously you don't have a clue.

pixel.png

Too bad... You still don't understand...
All my rectangles have the same size. You might have noticed. One rectangle represents one pixel. Look at the original. It's 2*2 pixels.
pixo.jpg

Now we want to display this picture on a 3*3 screen without black borders. If we use nice bilinear scaling, it'd look like this:
pixb.jpg

The upper left pixel is blue. The upper right pixel is blue, too. Half of he pixel in the middle of them belongs to the original upper left pixel, the other half to the original upper right. Both are blue so if you add 1/2 blue + 1/2 blue you get blue.
Now lets take a look at the lower left pixel. It's red like the original lower left pixel. But what about the middle left pixel? Half of it belongs to the original upper left pixel, half to the original lower left pixel. That's 1/2 blue + 1/2 red. Since you can't render it like that because you've only got one pixel to do it, it's averaged to violet.
In the middle of the 3*3 screen, all four original pixels would overlap. That'd be 3/4 blue and 1/4 red. Since you've only got one pixel to display it, it's rendered as violet but more blue than the violet in the middle left pixel.
The great thing about this is that every pixel counts equally. There is exactly as much percentual red influence in the lower picture as in the original. Also, everything is exactly in the right place. The disadvantage is that it looks a little blurry.
It's ok if you don't like the blurryness. But you have to stop talking crap like that there is pixel doubling involved. There is not.


To clarify: This is the quote I want you to take bacK:
DaveC posted on Oct 23 2005 at 06:58 AM said:
Since you can't do half a pixel some are doubled while other pixels are not giving an uneven look to the pixels especially when scrolling. Then sometimes a filter (blur) is added to try to hide the distortion, then you just have a mess.
That's bullshit. If you admit that, then I'm happy.
 
Last edited by a moderator:
Yeah, DaveC, I'm sure you never use interpolation on your computer's emulators. What I showed was a simple graph of what interpolation does.
 
A filtering hasn't to be a such bad idea. It's all depends on an actual aplication. Keep in a mind a TV does blur images too and adds many artifacts (not ideal convergence of RGB's electron beams). There also a situation when you can have a filter with a threshold which doesn't blur an edges.
 
Using OrR's example

You openly admit that there is blurring.
The disadvantage is that it looks a little blurry.
To DaveC, that is a distortion of colour since it induces colours that were not original there. The bilinear scaling is a way of avoiding the doubling of pixels but does so at the cost of introducing new colours to the image. (Colour distortion)

If you wanted to avoid inducing any other colours, how would it look? Selective pixel doubling. (Image distortion)

DaveC doesn't like either, others can live with one out of two.

edit: When I said 'To DaveC', I meant 'From DaveC's point of view'
 
yaustar posted on Oct 23 2005 at 07:01 PM said:
Using OrR's example

You openly admit that there is blurring.
The disadvantage is that it looks a little blurry.
To DaveC, that is a distortion of colour since it induces colours that were not original there. The bilinear scaling is a way of avoiding the doubling of pixels but does so at the cost of introducing new colours to the image.

If you wanted to avoid inducing any other colours, how would it look? Selective pixel doubling.

That was the original point. With a screen that is of the resolution and size of the GP2X/GP32 you can have it at 1:1 without all of these compromises. That gives the sharpest image. The border that may be there is small enough to not be a big deal.

As soon as you start stretching you have to make compromises. Now you have to either have selective pixel doubling which looks ugly, or a blurry image which looks ugly to some as of us well.

If the resolution is high enough you can do full doubling which isn't bad. Look at the Lynx emu for that. It is true 2:1.
 
Last edited by a moderator:
OrR posted on Oct 23 2005 at 07:22 PM said:
Ok at I'll let that count as taking the above quote back. Glad that has been resolved. <_<

This is the quote that I don't like:

"No, you fucking idiot, there is absolutely NO distortion."

When in fact there IS distortion (blurr).

This shows all from your Chrono pics.

scaletext.png
 
Last edited by a moderator:
DaveC posted on Oct 23 2005 at 07:41 PM said:
OrR posted on Oct 23 2005 at 07:22 PM said:
Ok at I'll let that count as taking the above quote back. Glad that has been resolved. <_<

This is the quote that I don't like:

"No, you fucking idiot, there is absolutely NO distortion."

When in fact there IS distortion (blurr).

This shows all from your Chrono pics.

scaletext.png

I'm with DaveC on this one. I definitely like the non-distorted look better. But, if you absolutely must make it something other than 1:1 or 2:1 then I think that what OrR is talking about is much better than simply every other doubling.
 
Last edited by a moderator:
First I agree it is the best to have a native resolution to any scaling. However not always this might be possible and filtering does have a merits. You are discussing a simple pointsampled and bilinear based filters. Indeed they are a crap. But there are far more sophisticated scaling algorithms we shouldn't forget about.

For an example look at the UQM game. It's a port of StarControl2 which originally runs at 320x240. It does offer four different methods of scaling: pixel doubling, bilinear, adaptive bilinear and an advanced biadaptive.

And finally we have something like the 2xSaI.

Now tell me if you still think a scaling must be sucking so much. Imho it doesn't have to as the examples above are showing.
 
Radek posted on Oct 23 2005 at 08:47 PM said:
First I agree it is the best to have a native resolution to any scaling. However not always this might be possible and filtering does have a merits. You are discussing a simple pointsampled and bilinear based filters. Indeed they are a crap. But there are far more sophisticated scaling algorithms we shouldn't forget about.

For an example look at the UQM game. It's a port of StarControl2 which originally runs at 320x240. It does offer four different methods of scaling: pixel doubling, bilinear, adaptive bilinear and an advanced biadaptive.

And finally we have something like the 2xSaI.

Now tell me if you still think a scaling must be sucking so much. Imho it doesn't have to as the examples above are showing.

Don't forget though these are being used for computers that are scaling in even multiples such as 2:1, 3:1 etc. 1.5 :1 would give different and not as good of results.
 
Last edited by a moderator:
DaveC posted on Oct 23 2005 at 10:59 PM said:
Radek posted on Oct 23 2005 at 08:47 PM said:
First I agree it is the best to have a native resolution to any scaling. However not always this might be possible and filtering does have a merits. You are discussing a simple pointsampled and bilinear based filters. Indeed they are a crap. But there are far more sophisticated scaling algorithms we shouldn't forget about.

For an example look at the UQM game. It's a port of StarControl2 which originally runs at 320x240. It does offer four different methods of scaling: pixel doubling, bilinear, adaptive bilinear and an advanced biadaptive.

And finally we have something like the 2xSaI.

Now tell me if you still think a scaling must be sucking so much. Imho it doesn't have to as the examples above are showing.

Don't forget though these are being used for computers that are scaling in even multiples such as 2:1, 3:1 etc. 1.5 :1 would give different and not as good of results.

Yeah, no a magic here. Yet still results will be better if something more advanced than a bilinear is being used. It's easy to check (I was using the Gimp2) that at 1.5:1 ratios (or any other) the bicubic filtering is giving a far better images than a plain bilinear. And there're also an oversampling and tricks with a LCD's subpixels to take advantage of.

My point is a scaling even in the bad situations (1.5 ratios or worse) can be tuned to be good enough to be passable.

Is there any handheld capable of something better than a simple bilinear filter (without taking out all of cpu's cycles of course!)?
 
Last edited by a moderator:
Radek posted on Oct 23 2005 at 03:00 PM said:
My point is a scaling even in the bad situations (1.5 ratios or worse) can be tuned to be good enough to be passable.


I think that sums up the whole argument here. Yes, those algorithms will certainly help make it look acceptable, but won't make it EXACTLY like the original.

Even simple double pixeled Snes emulator for Dell Axim 50v at 640x480 leaves something to be desired, especially the black outlines of bitmaps. Is it unacceptable? Hell no, but that's not the best I have seen, either.

There is no black magic here unlike those CSI episodes that the guy could zoom in infinitely to read fine prints.
 
Last edited by a moderator:
Back
Top