Slashdot Mirror
Pixel Inventor Goes Back To the Drawing Board
lawpoop writes "Russell Kirsch, inventor of the square pixel, goes back to the drawing board. In the 1950s, he was part of a team that developed the square pixel. '"Squares was the logical thing to do," Kirsch says. "Of course, the logical thing was not the only possibility but we used squares. It was something very foolish that everyone in the world has been suffering from ever since.' Now retired and living in Portland, Oregon, Kirsch recently set out to make amends. Inspired by the mosaic builders of antiquity who constructed scenes of stunning detail with bits of tile, Kirsch has written a program that turns the chunky, clunky squares of a digital image into a smoother picture made of variably shaped pixels.'"
Why are we suffering from it since so ?
I did not read the article, so I don't know if it's answered there.
Jeebus!! Invented the pixel. I'll be damned. :-P
Lost at C:>. Found at C.
Just exactly what the fuck is wrong with square pixels? They are the easiest to manipulate algorithmically.
I've abandoned my search for truth; now I'm just looking for some useful delusions.
I mean seriously..
In my early days of computing, pixels werent square.. they were rectangles (4:3 display, 320x200 resolution.. do the math)
Now who wants to write a rasterizer for non-rectangular pixels... any takers?
"His name was James Damore."
Somebody addressed this problem a long time ago, I see it on CSI every week.
Someone forgot to credit the Onion for this article.
Really, square pixels are an invention? Meh. And he might want to credit Georges Seurat for his new direction.
"National Security is the chief cause of national insecurity." - Celine's First Law
he used 36x the number of square pixels
Here's a relevant article about it: http://alvyray.com/memos/6_pixel.pdf
nevermind that most computer displays were using rectangle pixels, nevermind that the article claims there was only 1 programmable computer in the US in 1950
look at his closeups, what do you see? SQUARES, putting a bunch of squares in a new shape doesnt change the fact they are squares
He just "invented" JPEG too!
Another World did something similar to compress images -- building them out of shapes: http://media.giantbomb.com/uploads/0/5052/407468-another_world1_super.jpg
Seems these squares do just fine. And I suspect any issues with grids of squares can be fixed with higher and higher resolution grids. And manufactured sensors and displays will be using grids that will be uniform.
Now who wants to write a rasterizer for non-rectangular pixels
From the article: The pixels are still square; they're just cut into two pieces along a line through the pixel, and each piece has a color. (It sort of reminds me of S3TC.) The edge of a polygon would have one piece for the front and one for the back, and any other points along it would have one piece for each of two texture samples.
...but I couldn't get past all the square pixels on the page.
There's no -1 for "I don't get it."
He says that it was the logical thing to do, it still is! Keeping the pixels square makes sense in almost every computer science aspect I can imagine, and this guy have had 50 years of regret and came up with something that's comparably very hard to implement in scanners, memory, screens and software. Triangles and hexagons are two other ways he might have gone that's comparably simple, but squares are more intuitive. I think his contribution in the past was brilliant. He really should have no regrets.
- Henrik
- when the Shadows descend -
I would never have known that the Japanese have blurry genitals.
Eloi are stupid, throw morlocks at them!
This sounds like the ongoing debate between analog and digital audio. Everyone likes using images like these during the debate, but given enough resolution (bits), the closer the digital audio will be to its original analogue (electrical) source.
I'm god, but it's a bit of a drag really...
And that's not just because Civ5 is coming out soon...
Blar.
square
Square pixels are easy to manipulate to be sure, so are single-core CPUs easier to code against, but the world is not perfectly right-angled. I'm now waiting for my flying car; which will use nano-morphing, variable shaped pixel in-car & HUD displays, and all controlled by a hella-core processor system running Hellabuntu. The sky's the limit! Or something like that.
This is the NSA, we're gonna geet U h@x0r5! Also, what is a h@x0r5?
But maybe that's because that's what we used in early games.
It does translate fairly well to an interlace design for scans, though.
-- Tigger warning: This post may contain tiggers! --
That's what it sounds like to me.
Square? Not on my old Trash-80! Heady days indeed, populating that 128x48 grid of monochromatic pixels. The aspect ratio really screwed with attempts to generate graphics using trigonometry.
to fit a larger baby image on his 1957 computer?
From what I get from the article for him it is like a day has not passed since then and things like image compression were never invented.
Oh, wait, that is what Alzheimer's does to you... I see...
Anyway, we should thank him for the pixel (and for it being square and not something ridiculous that would give us problems for years to come...).
Violence is the last refuge of the incompetent. Polar Scope Align for iOS
Working in the web division of a semi-fine jewelry retailer.
graphic artist: Do you know computers?
me: I should hope so. Do you have a question?
graphic artist: Yes. It's with Photoshop.
me: Ok, I might be able to help. What's the problem?
graphic artist: Ok, let me zoom in here. You see what I have here? (zoomed in so that the pixels were big blocks on the screen.) Everything is really blocky.
me: Understandable at this view level.
graphic artist: Well, it's not working for me. I need to be able to get a smaller shape in here but it's all too blocky.
me: Let me get this straight. You want to get a shape in the image smaller than a pixel?
graphic artist: *beaming* Exactly! How do I do that?
me: I'll look into it. *slowly edged away*
And she was getting paid three times what I was. Things like this make me want to lock myself in the server room, trip the halon and wait for the blackness to take me.
Kwisatz Haderach
Sell the spice to CHOAM
This Mahdi took Shaddam's Throne
The configuration of the variable-sized "pixels" depends on the image, so you're not going to get a new screen with more detail -- this is for storing images. From what I can tell, he's doing a basic form of an old and well known compression technique ("macroblocks" in JPEG, H.264, and others) and calling it a new form of pixel.
in one, a seam divides the mask into two rough triangles, and in the other a seam creates two rough rectangles
My digital alarm clock already invented that.
I'm sure if the picture on the left was resized to match the pixel density on the left it would look just as good..
1) A pixel isn't "invented" by anyone. A pixel is just a concept that is so straightforward, like the wheel, language and adding numbers. It's not a question of which single person "invented" it. It's just a question of, once the technology is there, it WILL be used, no matter what.
2) What kind of screen are you going to use for that? Each pixel can have different types of pixel sizes so no screen could fit that. A square grid is the most uniform division of 2D space into units.
3) If this would have been about hexagonal pixels, I'd have found this cool.
4) At best, this is a new compression scheme for storing pictures - but certainly not a way to display them (see 2))
5) Non square pixels are not a new idea, see for example sensors of cameras.
Perhaps it is a clever algorithm, but the summary and article make it sound like he is re-inventing the pixel. I don't think that is correct. The example shown starts with a square-pixel image, and outputs another square-pixel image, just at a finer resolution with less blockiness.
While it does appear to work, it isn't clear to me how much information can be inferred correctly. Furthermore, cameras often don't use square sensors to begin with, so this isn't directly applicable to the raw image format.
Obligatory paper on the subject of pixels: A Pixel Is Not A Little Square, A Pixel Is Not A Little Square, A Pixel Is Not A Little Square! (And a Voxel is Not a Little Cube).
The current problem is that on an LCD display, the Red, Green, and Blue pixels are adjacent to each other, not co-located. Coming up with a scheme to make all 3 colors appear to emanate from the exact same point would be a useful development.
I've abandoned my search for truth; now I'm just looking for some useful delusions.
While it describes the algorithm, does anyone know of a reference implementation that I can play with? The results are quite compelling!
Slashdot's rate-of-post filter: Preventing you from posting too many great ideas at once.
From TFA: "...So one thing you can do is say, I’m going to assume the next pixel is like this one. Don’t talk to me, don’t tell me anything about the image, until you get something different..."
Brilliant! He's reinvented Run-Length Encoding!
"Remember when I said I would never lie? Well, that was the first time."
We just got the first 300 plus dpi screen in iPhone 4, and obviously that will come to other devices. At that point, you don't care about the pixels anymore because you can't see them. What is onscreen does not appear to be made out of pixels, it appears to be made out of curves and lines. Most photographic prints are less than 300 dpi. Most people have never seen an image with higher resolution than an iPhone 4 screen. We don't talk about the shape of the pixels in a chemical photo print, we talk about the image that is shown there. We're entering that time in onscreen images now also. As great as pixels are, they are better when you can't see them.
If you have non-square pixels and want to create horizontal or vertical lines, you will get the exact same problem as we currently have with square pixels and diagonal lines.
The pixel element simply can't have a shape that naturally supports lines in all different directions.
Many image-enhancement techniques exist that do just this, and this is not really new. In fact this proves that square pixels work just fine to transmit the information, but the image can be enhanced to a larger resolution by non-linear techniques that work better than simple [traditional] upsampling.
Help! I am a self-aware entity trapped in an abstract function!
The printing industry has dealt with this sort of thing for a while. Read up on Stochastic Screening. Not quite the same, but it gives you a sort of idea of the problem of mapping continuous tone data in a non-continous space.
Are easy to grasp, and have all sorts of uses. You can produce them on a CRT simply using a scanline technique, they're simple to address logically, and sections can be moved around (albeit by discrete amounts) trivially, and you can apply some simple but useful filters very easily.
Other representations in addition are also useful. Some people find fourier transforms of the data helpful. Others find it useful to identify lines and discrete shapes.
It seems that Kirsch has been playing with another representation. Seems to do a rather good job of enhancing certain types of typical real world image. Good stuff, but not a reinvention of the pixel.
http://nvl.nist.gov/pub/nistpubs/jres/115/3/V115.N03.A03.pdf
Sometimes they fool you by walking upright.
A Pixel Is Not A Little Square, A Pixel Is Not A Little Square, A Pixel Is Not A Little Square! (And a Voxel is Not a Little Cube)
First, here's the actual paper, since it clarifies what exactly he's suggesting and doesn't seem to be linked anywhere in the article.
It's not a suggestion that we start using non-square pixels for displays or cameras or scanners or what not, though he's certainly not being very clear about anything and the reporting on this is just making matters worse. What the paper proposes is a method where:
1) The image is split into 6x6 blocks
2) For each block, you go over the four rotations of the two following two-section masks:
The triangular mask:
ABBBBB
AABBBB
AAABBB
AAAABB
AAAAAB
AAAAAA
The rectangular(ish) mask:
BBBBBB
BBBBBB
BBBAAA
AAAAAA
AAAAAA
AAAAAA
for a total of eight effective masks, and average the values under each section, resulting in two values, A and B.
3) For the mask and rotation that has the largest difference between A and B, you output the mask, the rotation, and the A and B values, resulting in 19 bits from a 6x6 (288 bits) block.
Though he talks of non-square pixels and whatnot, it's really just a compression algorithm. A really stupid one. Basically it's a bad variation of vector quantization, with lots of baffling details. Why 6x6 blocks? Why those specific masks? Why are you maximizing contrast instead of minimizing error like any sane person would do, WHY? There's no rationale given for any of these choices, not theoretical, not empirical, not even subjective.
The same sort of rigor extends to his comparison, where he compares his compression algorithm to, instead of, say, another compression algorithm, the image apparently simply downscaled and then scaled back up. And not even with a halfway decent resampling algorithm, but with nearest neighbour. Not to mention that the "non-square pixels" version has 2.375 times as many bits to work with. If he'd done a comparison to a reasonably modern compression algorithm like JPEG, the results would be much less favorable to him.
tl;dr Some old guy put together his My First Compression Algorithm kit and it's being treated like a revolution in graphics by ignorant reporters. Nothing to see here, move along.
So he just drew pixels with even smaller pixels? Can we make something to redraw the pixel pixels with even smaller pixels that split them up. Like some sort of megapixel.
Having variable size pixels wouldn't be so bad for low power devices. You could essentially render images with fewer pixels and thus reduce the amount of screen space to refresh. This would save power (and with the right algorithms, cpu time as well).
The algorithm he created looks a lot like HQX which is used mostly to scale old video games. His algorithm seems generalized to work on high-color images while the HQX algorithms expect something closer to 16-color or 256-color images. HQX probably deals with dithering better.
The whole thing about "square pixels" is just the media angle.
You're still a square, and the Summer of Love is over!
I wonder if anyone at Pixar actually uses a "drawing board", and if they do if there's any benefit to going back to it.
You are welcome on my lawn.
Just to elaborate on this, you need to distinguish between a square sampling grid, and a square (box) reconstruction filter. An image is really just a set of samples, the color at a grid of points in the image, not a grid of the average color over a square area (not usually, at least). It's very similar to a PCM sound file, where it's a recording of the instantaneous amplitude at each point, not the average over the sampling period. If you reconstruct the sound wave with a box filter, you get a crappy result. The same with images; a gaussian reconstruction filter is usually better than a box filter. LCDs are actually a step backwards in this regard, at least for images, since they have much more boxy elements than a CRT.
Showing my ancient history, I remember an early presentation inside DEC for what became the DEC 1M pixel terminal that became the basis for most of X11's design, from perhaps 1982. The pixels were only 1 bit each (ie black or white). They were debating a design where the scan lines were .717 as far apart vertically as the pixels were horizontally and each odd numbered line was shifted horizontally (by the CRT scanning circuitry) by 1/2 pixel. The result was a hexagonal layout that also nicely filled the 4x3 screen while the source was a 1024x1024 array of bits.
Inability to nicely draw vertical lines is what killed this idea. In the end it was more practical to lose 256 lines of there expensive display memory off the bottom and make square pixels. This extra display memory ended up being used for off-screen images, it's availability but limited size also greatly affected the design of X11.
They also had ethernet cables that were the really thick yellow things with the vampire tap that you screwed onto them to add a new machine to the network. All of this was in a the Dec research department showing the future of computers...
That is what it looked like to me as well, but I found the actual paper, and he is creating his "non square-pixel" image from a larger image, not upscaling it from a smaller one. In other words, it is basically just a form of poor-man's compression where you replace each 6x6 block with one of 8 decompositions containing two coefficients each.
Really, back to basics is bottomless if you chose to look beyond convenient hype.
``Tension, apprehension & dissension have begun!'' - Duffy Wyg&, in Alfred Bester's _The Demolished Man_
His description of the algorithm (and the processed image) reminded me very much of the scalers commonly used in video game emulators such as MAME and almost all (S)NES/Sega emus. They're usually named SaI, Eagle, AdvMame etc.
They basically look at a 2x2, 3x3 or 4x4 cluster of pixels, figure out any high-contrast boundaries and try to intelligently interpolate according to those details.
The expensive version of this is known as "fractal interpolation". It is more CPU intensive Mr Kirsch's method but the idea is the same: convert the image into contrast vectors then rasterize back to a higher resolution. His algo and the ones used in emulators are merely pixel-based approximations of the fractal approach.
Okay. Minus five points from the "graphic artist" for not knowing how to resample the image. Plus one point for trying to improve her knowledge instead of suffering in silence.
Minus one point from you for not knowing something that's not directly in your field.
Minus ten points from you for not even trying to help.
Minus fifteen more points from you for being a jerkass about it on Slashdot.
So she's down four, and you're down twenty-six.
81 according to the article, that's how old he is.
Most of you youngsters in here maybe don't get it, but as I have now blown through my teens, my twenties, and now most of my thirties I have definitely begun to acquire more and more of a distaste for that which I do to make money.
If I make it to 81 and I still give two craps about technology in general, much less have enough enthusiasm left to try and contribute, then I'll consider myself an extremely lucky man.
Kudos to this guy for working in tech SINCE THE FIFTIES and not burning out.
How many of us will be able to say the same?
regular pentagons cannot be tiled,unlike equilateral triangles squares and regular hexagons
"An image is really just a set of samples, the color at a grid of points in the image, not a grid of the average color over a square area (not usually, at least)"
No sample, whether a pixel or a sound sample, is EVER a point sample. It's NEVER an instantaneous value. That would require that your sampling function be a delta, and delta functions have nice little features like taking infinite values and having infinite bandwidth.
A real sample is always an average over some finite time or space (or other dimension), weighted by the sampling function. In many cases that sampling function is indeed designed to be as close as possible to a rectangle (or boxcar) function.
Big deal.
September 2011: Looking for Cocoa/iOS work in Boston area Cocoa Programmer Quincy, MA
Fuck Russell Kirsch. I love square pixels! I would give anything for all the data I work with to be in square pixels. I have wasted literally years of my life dealing with non-square pixels in the video industry. Fuck that. If there were 2 things in my job I could get rid of at the snap of my fingers, they would be non-square pixels and interlacing.
The approach looks very much like wedgelet approximations see theory and images at http://ibb.gsf.de/homepage/laurent.demaret/wedgelet/docu/wedgelet_short_guide/node1.html
Non square pixels are not a new idea, see for example sensors of cameras.
Alvy Ray Smith, computer graphics researcher and co-founder of Pixar, will tell you that A Pixel Is Not A Little Square, A Pixel Is Not A Little Square, A Pixel Is Not A Little Square! (And a Voxel is Not a Little Cube). I don't necessarily agree with everything in the article (e.g. coordinate systems) but it is worth a read.
It is probably worth referencing Alvy Ray Smith's (co-founder of Pixar) enlightening memo from 1995 entitled: A Pixel Is Not A Little Square, A Pixel Is Not A Little Square, A Pixel Is Not A Little Square! (And a Voxel is Not a Little Cube)"
First world problem...
Hexagons.
Agreed, but the voronoi diagram of those sample points gives the squares.
By the way, I forgot to mention the article is very interesting, thanks for the link!