Why are square shaped pixels so widely used when triangle shaped

why do retarded people make new threads when we have an ongoing stupid questions thread?

>>55029030
Because it is not a stupid question.

At least on a computer, most things are going to be shaped around straight edges. I wont deny that the triangles look better in that pic tho for organic structures.

>>55029024
If you can figure out a way to make the LEDs in a screen triangles you have my permission to make millions of dollars off of this.

>>55029037
It's a bait thread. The left side has a much higher pixel density -- you see that "triangle" at the tip of the ear? That's a bunch of pixels in 6 times higher resolution than the left side. It's just an effect applied to the original image.

>>55029024
>what is chroma subsampling

Check out >>>/sqt/

>why does it look better if I sextuple the resolution?

>>55029037
maybe it's not stupid in the sense that it transcends stupid and approaches fucking retarded.

>>55029062
>If you can figure out a way to make the LEDs in a screen triangles

Modern /g/ tech knowledge

>>55029024
because that's literally made of of smaller square shaped pixels

>>55029104
Obviously they can be made into dots and placed into a triangular configuration but I would assume you would need to write entirely new drivers to tell the computer how to output to such a different design.

I just woke up and am not thinking in straight lines yet, bite me.

>>55029143
>this question is stupid
>no it's not
>here's one obvious problem
>okay so i'm still half-asleep 10 minutes after posting the thread, everyone cut me some slack!

>>55029062
>If you can figure out a way to make the LEDs in a screen triangles
some oled screens have circle/round/diamond shaped subpixels, but im sure thats not what op is trying to ask

>>55029024
A "triangle-shaped" pixel does not exist. Screens are physically made of grids of _square_ pixels, and if you zoom in enough on something you will see that.
Too lazy to link you to the SQT but that is where this belongs.

>>55029167
the OP literally just woke up and posted on 4chan. he's "not thinking in straight lines". i think it's safe to bet that the OP didn't mean anything in particular because he wasn't thinking anything in particular.

>>55029024
On an LCD pixels do their job perfectly. Unless you break out a magnifying glass or microscope they're all just gonna look like dots anyway.
not to mention the triangles in the right comparison are still made of smaller square pixels

>>55029155
I don't know know why you think I'm OP. I was sarcastically saying OP has my permission to make millions because obviously someone has thought of this before but it wasn't economical enough to make money off of. It's a joke, and now that I've explained it to your oblivious ass it is no longer funny.

Good job, ass.

>>55029204
ITT: /g/ finds a way to make millions of dollars by reinventing the pixel.

>>55029232
I'll make the logo.

>>55029204
oh shit dude did i hurt your feelings? i'm really sorry. i basically only come to 4chan to validate your hapless attempts at humor and generally shitty posts, so i'm feeling really bad about this whole incident right now.

>>55029244
>gets pointed out how idiotic he is
>uh-hu y-YOU MAD yeah, that's it

Loving
Every
Laugh

I came here to shitpost and watching you squirm is great. Thanks for the wake-up faggot, better than black coffee.

File: WebP_Test_mini.png (2MB, 1011x3032px) Image search: [Google]

2MB, 1011x3032px

Because it doesn't matter anymore. Phone displays have very high PPI so whether they use triangle or square pixels means jack shit. This also applies to monitors now that they have exceeded 100 PPI (see 4K monotors).

What does matter is what type of lossy encoding is applied to pictures if any. JPG leaves behind blocky artifacts that destroy detail and and colors for the most part. Webp does not, even at -q 0, the lowest quality possible pictures still retain a good amount of detail compared to JPG.

>>55029268
i'm sure the embarrassment of being outed as a fucking retard is a stark wake-up. if it'll help you start your day and leave 4chan for a few hours, i'm happy to out you.

>>55029311
>being outed
>on an anonymous imageboard

I don't think you know how 4chan works. Don't worry, I'm sure you'll get it down pat before you have to go back to school. We all remember our first summers here.

>>55029311
>>55029244
>>55029155
dubs checked checked checked. BTFO'd

File: 1465674275828.jpg (99KB, 782x566px) Image search: [Google]

99KB, 782x566px

>>55029024
It is much easier to make square pixels, than triangular ones. So why would you use triangle shaped pixels if you could just increase the resolution?

File: 1363155549790.gif (80KB, 170x207px) Image search: [Google]

80KB, 170x207px

>>55029244

You should compare square pixels to triangular pixels with the pixel density compensated for.

Doesn't the picture with the "triangle" pixels essentially have double the pixels?

File: dumbass.png (438KB, 1000x1000px) Image search: [Google]

438KB, 1000x1000px

>>55029290
or just use glorious png

>>55029143
>Obviously they can be made into dots and placed into a triangular configuration but I would assume you would need to write entirely new drivers to tell the computer how to output to such a different design.
>I just woke up and am not thinking in straight lines yet, bite me.

But you think the pixels in an lcd panel are leds

You are a fucking retard

>>55029360
>Protip: you can't
At least not without a microscope but by then it doesn't matter anymore.

>>55029375
>lossy PNG with clear grainy reduced color dithering and color banding
fuck off

>>55029375
>lossy PNG
is this nigga serious...

>>55029337
What about some kind of new panel technology that can dynamically change its pixel shape?

File: webp is worse than png.png (79KB, 1119x608px) Image search: [Google]

79KB, 1119x608px

>>55029290
sub pixel format matters a lot, its one of the pertaining causes of low viewing angles, and on a phone that matters since its not always on you hand

image encoding means nothing on the issue at hand, since data gets fed raw to the display
if anything, saving images on broad compatible lossless compressible formats makes them easier to display on any device at hand

>>55029377
>all screens are LCDs

Okay, idiot.

Completely ignoring LED/OLED, plasma, CRT. etc. there I see.

If you're going to try to call someone a retard you should probably make sure you're not a retard too.

>>55029439
Hey look, it's lossypngfag.

>>55029439
>lossy png still bigger than a 24-bit RGBA Webp
Impressive

>>55029430
But what would be the advantage?

>>55029430
What about it? Does it exist? No? Cool. Shit theory.
Eventually triangles lose their benefit.

>>55029460
>>55029474
>running out of excuses
0.4% fellas, 0.4%...
oh and 69% worse, btw

>>55029456
>Completely ignoring LED

And do you think those displays are using LEDs as pixels?

Because they are LED BACKLIT LCDs

Unless you were talking about giant roadside billboards, which actually are a bunch of leds used as pixels.

Your a fucking idiot and emblematic of how clueless most of the people on this site are about the most basic aspects of tech.

>I meant oleds when I said led!!!!!

File: amoled-vs-lcd-wide-viewing-angle.jpg (27KB, 450x293px) Image search: [Google]

27KB, 450x293px

>>55029439
You're late to the party. The viewing angle problem has already been fixed by amoled.

>>55029527
see >>55029167

>>55029430
...you mean a vector graphic?

>>55029456
Literally kill yourself, you have no idea what you're talking about.

>>55029576
>...you mean a vector graphic?

not at all how vector graphics work

another clueless fuck itt

>>55029576
Something along those lines.

>>55029430
a phosphorus monochrome crt display (like the ones in analog oscilloscopes) have no/infinite pixels

>>55029024
who is that smug bastard?

>>55029638
Yes, but the goal here would be to actually improve image fidelity/quality

>>55029430
Jesus christ just drop the idea already, it sucks. Also keep in mind how content will have to be converted to look right with a different pixel shape.

>>55029659
increase ppi

If it was possible, at a reasonable cost, it'd have been done already.

See, for example, old vector displays.

>>55029675
>Also keep in mind how content will have to be converted to look right with a different pixel shape.

More like all the fucking issues of converting images captured by a bayer pattern sensor to triangles

there are already enough problems with that shit

>>55029024
Squares are easier to work with for a dozen reasons you should be able to realize on your own.

>>55029024
>Why does it look better if I dramatically increase the resolution and then also encode it with fewer artifacts?

Do a proper comparison you faggot

File: JC2.png (1MB, 1280x720px) Image search: [Google]

1MB, 1280x720px

>>55029508
More like 30%

https://my.mixtape.moe/dygfwj.webp

>inb4 you turn my PNG into a lossy one full of reduced color grain and color banding

>>55029721
That's fucking simple though. Take the existing stuff and make it use 2 triangles and treat them as 1 square.

>>55029799
if squares are easier to work with then why does every single GPU and rendering engine convert squares to triangles first?

>>55030427
>That's fucking simple though. Take the existing stuff and make it use 2 triangles and treat them as 1 square.

Then there is literally zero point, because now you've made square pixels

bravo

>>55029290
This. Fuck JPG.

>>55030442
>if squares are easier to work with then why does every single GPU and rendering engine convert squares to triangles first?

you mean quads to triangles?

because modern gpus dont handle quads as well as tris

this wasn't always the case though

Also

polygons =/= pixels fucking retard

>>55030473
>triangles != triangles
okay then

>>55029414
>lossy PNG

File: 208293_012.jpg (47KB, 320x240px) Image search: [Google]

47KB, 320x240px

>>55029024

>>55029375
>438 KB

No thanks

>>55029024
The answer is because mathematically too complicated. Square pixels are much better suited for algorithms (you can easily calculate most things in cartesian coordinates and it's very easy to convert between those pixel coordinates etc).

>>55031372
>"pngquantis a command-line utility anda libraryfor lossy compression of PNG images."

https://pngquant.org

It basically reduces colors and adds dithering to the ones left over. It always looks like shit at 100% zoom and is still bigger than a Webp lossless image.

File: diff.png (8KB, 1280x720px) Image search: [Google]

8KB, 1280x720px

>>55030346
>complaining i might post a "lossy" png
>posts a lossless webp himself
are you really that desperate?
oh well, here is your (you), your
>webp is 0.4% better then png
and your
>webp is 69% worse than png

>>55031666
fuck, that lossless is missing a couple of quotes around it
>"""lossless""" webp

>>55029024
it's typically easier to use more of a simple thing than less of a complex thing

how would you make a display that uses irregularly-shaped pixels?
would the additional encoding/decoding effort be worth using it over simply using more square pixels?

>>55031459
That's not the answer. The maths wouldn't be too difficult. And you only need to do them once.

File: 1374460484248.jpg (210KB, 555x555px) Image search: [Google]

210KB, 555x555px

okay, so webp is apparently better. Now, a simple question to all enthusiats: can it reduce the size of a file by a factor of 2-10 preserving quality? ANY practical, sustainable, useful way? If not, then why even bother?

>>55031996
The math is simple: you need to triangles to make a square. Double your data size?

>>55032162
Fuck off weeb

>>55032162
>webp is apparently better
0 . 4 %
.
4
%

>>55031996
Literally every algorithm in image processing would need to be changed. Image scaling, all kinds of rendering (bezier curves, fills etc), all linear operations (sharpening, blurring etc) would be a disaster in that weird coordinate system and so much more.

Literally every algorithm would become much slower. The matrices as they are used today have a very good reason, believe me.

>>55032162
Webp:
>has lossy encoding
>has lossless encoding
>supports transparency
>supports animations
>is 25-50% smaller than a non-lossy PNG, see >>55031490
>has about 80% web browser support
>is 50% smaller than a typical JPG at the same visual quality.

It's literally the best general purpose image codec that can exist as of right now.

File: LCD_TFT_Screen_Closeup[1].jpg (947KB, 999x670px) Image search: [Google]

947KB, 999x670px

>>55029186
No they're not, get a magnifying glass and see

This is why good renderers have so-called "subpixel hinting"

>>55032312
>No they're not, get a magnifying glass and see
Those looks pretty square to me anon

Rects are faster to calculate than triangles.

File: 1378640770281.gif (32KB, 250x250px) Image search: [Google]

32KB, 250x250px

>>55032191
>"I got nothing" -- the post
kys
>>55032290
finally, numbers. so we're talking about a factor of 2 compared to JPG, a factor of ~1.3 for PNG.
On what testing set? I don't know benchmark standards for such things, just to understand how realistic these are. I mean images of random noise for example would be rather unrealistic.

>>55032368
Rather, individual points in a 2D array are better than calculating triangle positions.

>>55029024
>>>/vr/

File: daphuc.jpg (119KB, 602x655px) Image search: [Google]

119KB, 602x655px

>>55029024
>pic related

File: SquareAndRectangle[1].jpg (39KB, 676x277px) Image search: [Google]

39KB, 676x277px

>>55032358
It's ok, it's not always easy to tell shapes apart.

In this case, the elements are rectangles with a ratio of 3:1, and each element is only one of red, green or blue.

>>55032392
see >>55030346 and >>55029290

Don't feed the lossy png troll in this thread.
Lossy PNGs do exist btw, see >>55031490

Webp is our next JPG, PNG, and GIF all in one package. If it wasn't any good sites like facebook and ebay would not use it.

>>55032487
Those are subpixels not pixels you retard.

>>55032312
Can someone explain to me how subpixel rendering works for coloured text? With black text any subpixel can be set to black to add extra resolution.

If you had say pure green text though there's only the green subpixels can be used to add extra resolution. If you've got an RGB layout the green subpixel is in the centre and adding another subpixel would just be the same as adding a whole pixel.

How come coloured text still looks as smooth as black text and isn't jagged?

File: a.png (4KB, 43x39px) Image search: [Google]

4KB, 43x39px

>>55032487
this is one pixel

the individual colors are sub-pixels, and is simply an artifact of the display technology (different display types will have different sub-pixel arrangements, with some even overlapping them to make "true" single pixels)

>>55029024
Because in your image you've got two triangles in the space of each square pixel. You've basically doubled the resolution.

Half of those "triangles" aren't even triangles anyway. The big pixels seem to be split up in a whole bunch of weird ways. That wouldn't work for a display. You're better off just doubling the resolution to get the same effect.

File: LCD-rgb-subpixel-matrix.jpg (11KB, 250x245px) Image search: [Google]

11KB, 250x245px

>>55029186
>Screens are physically made of grids of _square_ pixels

No they're not

>>55032503
>>55032560
That is three subpixels. You can treat them as one pixel, with slightly shittier results than normal.

On the close-up of the LCD, you can see how sequences of 3 pixels are not naively being treated as a single pixel. Sometimes, groups of 4 are used.

>>55032497
okay, fair enough. Thank you for the info.

>>55029527
>if you want privacy dont use amoled

File: 480px-Pixel_geometry_01_Pengo.jpg (64KB, 480x480px) Image search: [Google]

64KB, 480x480px

>>55032673
typically pixels are addressed, processed, and sent to a display as a complete group of the 3 primaries (RGB)
sometimes the subpixels of a display type are exploited (such as LCD subpixel-based font rendering), but these are exceptions, and are handled at a higher level

>>55032532
You can easily transform text that is (0, 0, 0) (black) on (1, 1, 1) (white) (I use floats for convenience) to any color (r, g, b) on (r', g', b'). It's a simple linear operation that is determined by this property:

m * 0 + b = r
m * 1 + b = r'
=> b = r, m = r' - r => f_R(x) = (r' - r) * x + r.

And you just apply that to every subpixel. Pretty simple, really.

File: R.png (2KB, 500x542px) Image search: [Google]

2KB, 500x542px

>>55032532
>there's only the green subpixels can be used to add extra resolution.
You don't just use the green subpixels. You just need the average to be green over a small 2-3 pixel area, and it'll blend together.

>>55032577
Well, yes and no.
Subpixels form square pixels, so he's not entirely wrong in saying that a monitor is made up of a grid of square pixels.
Saying that a monitor is made up of rectangular subpixels would also be correct.

>>55030545
Pixels aren't triangles you idiot. They are used for thr display, they can only be used by the monitor and altered by the computer. The polygons are 3d triangles that are rendered and processed by the gpu/cpu and all thosr microprocessors inside them, they aren't pixels, they aren't display, they are converted into pixels to be viewable.

>>55032857
From a technical and physical point of view, that's equivalent to arguing that monitors are made up of 160x200 2:1 rectangular superpixels like on a C64.

The monitor is absolutely explicitly designed to allow turning on individual subpixels. Yes, it's useful to treat 3 as a full color point. Yes, it's useful to treat 12 as a full color point with lower resolution. That doesn't change the physical layout of the screen.

>>55032765
A monitor doesn't address or process single pixels. But like you say, yes, each individual subpixel is individually set in about a byte's worth of data on the display stream.

Often groups of three are closely related due to the simplification of dealing with one RGB point rather than multiple R, G and B points.

File: subpxl.png (35KB, 914x550px) Image search: [Google]

35KB, 914x550px

>>55032532
Or, here's another attempt:

When you subpixel render an image that is grayscale, you basically do the following: You render the image with three times the width of the final image. Then you downscale the image and assign blocks of three pixels to the subpixels of the resulting image, meaning (pixel1, pixel2, pixel3) --> (pixel1_R, pixel2_G, pixel3_B). When you want to do the same thing with a colour image, nothing changes, the only difference is that you assign the corresponding colour channels:
(pixel1_R, pixel2_G, pixel3_B) --> (pixel1_R, pixel1_G, pixel3_B). What you get is a colored subpixel rendered image.

Pic related. How three successive pixels are assigned into the final image.

>>55033023
>A monitor doesn't address or process single pixels.
i didn't say they do, though modern digital displays would need to do something

what i'm talking about is for example, VGA has pins for R, G, and B seperately, and a pixel is sent simultaneously to the display, not one color at a time. digital display interconnects might be different (it's probably simpler to push the RGB values serially at 3x the speed than split them into 3 wires)

yea, you can ignore the concept of colors and treat the display as being 3x the horizontal resolution, but unless you're trying to exploit a display types' subpixel layout, the smallest element you or your program cares about is a pixel

>>55032312
>good renderers
If you think subpixel hinting looks good you need to get your eyes checked

enjoy your rainbow text

>>55029243
4 hours and still no trixel logo. What the fuck are you doing with your life

File: a.png (18KB, 655x999px) Image search: [Google]

18KB, 655x999px

>>55033287

File: a.png (18KB, 655x966px) Image search: [Google]

18KB, 655x966px

>>55033452
oops, wrote the wrong thing

>>55032995
>The polygons are 3d triangles that are rendered and processed by the gpu/cpu and all thosr microprocessors inside them, they aren't pixels, they aren't display, they are converted into pixels to be viewable.
The details are irrelevant. You claimed triangles are harder to work with than squares. I countered by asking you to explain why triangles are easier to work with than squares in graphics processing.

>>55032995
>>55033597
It's ironic because all of the algorithms that mathematicians invented to deform 3D meshes (aka the only ones that matter) are designed to operate on roughly box-shaped quads rather than tris or anything else.

>>55033597
What the fuck do you think trigonometry and calculus are? Is this a serious question or are you shit posting? Decades of research into calculating triangles are the reason they're easier to work with.

>>55033650
Yes, and they probably also used real and complex numbers in the process. But then you try implementing it in reality and need to come up with encodings like IEEE to embed an approximation of the complex/real result as binary logic, since that's what computers prefer.

Same reason you need to transform all of those pretty quad-based algorithms into ones that will work on triangle primitives, since that's what computers prefer.

>>55033665
>Decades of research into calculating triangles are the reason they're easier to work with.
Then explain this statement:
>Squares are easier to work with

It seems like you're directly contradicting yourself at this point.

>>55032392
>kys
Kill yourself.

>>55029024
Are you on drugs?