20

I'm confused about the term "pixels per inch"

iPhone 6 has 326 ppi

there's 25.4 millimeter in 1 inch which I can barely see... how much more of 326 pixels inside 1 inch....

So does Pixel have a standard size?

  • 48
    Does 10% have a size? – blankip Jul 22 '16 at 16:50
  • 20
    Lol, I'm guessing you've never seen many screens in your life. Contrast an old 17" VGA monitor to a new 10" Retina display or whatever. Clearly all pixels are not the same size. – underscore_d Jul 22 '16 at 16:53
  • 23
    that question makes little sense. "pixels per inch" means just that. That's the number of pixels in a linear inch in both directions. 326 ppi means the pixel on the iphone screen is 0.078 mm wide. – njzk2 Jul 22 '16 at 17:33
  • 31
    My phone's display is 1080x1920 pixels. So is my monitor's. – Aza Jul 23 '16 at 0:49
  • 11
    All three of these have 1920x1080 px screens. – gronostaj Jul 23 '16 at 20:14
64

The size of a pixel varies by device. For example a 24" monitor running 1024x768 has bigger pixels than a 20" monitor running the same resolution.

The pixel is the smallest dot that can be turned on or off (or given a colour). The higher the resolution (more pixels) for a given display size, the smaller the dots (pixels) and the finer the detail that can be displayed.

  • 36
    Not quite correct. A subpixel is the smallest element that can be turned on or off. On a LCD, there are 3 subpixels: red, green, and blue that make up a pixel. These 3 subpixels get turned on and off to create a pixel color. Fix the answer and Ill upvote. – Keltari Jul 22 '16 at 16:52
  • 31
    @Keltari Well if we're nitpicking that's also not entirely correct, since you don't have to have 3 subpixels per pixel. PenTile uses RGBG for example, so you only get 2 subpixel per pixel. Nothing as simple as it used to be, not sure how you'd sensibly measure these non standard layouts. – Voo Jul 22 '16 at 18:08
  • 8
    @Voo Not correct: RGBG is four pixels which all contribute to one image pixel. Further, there are plenty of algorithms which "share" the G-pixel info across adjacent 2-by-2 subimage elements. The MTF of digital color image sensors is a tricky thing indeed. – Carl Witthoft Jul 22 '16 at 18:30
  • 33
    But, how many atoms does a pixel have? – Matthew Jul 22 '16 at 20:11
  • 2
    @Carl A single colored (say red) dot is a subpixel, with RGBG having 4 subpixels (2 green, red, blue). But those are addressed as 2 pixels, not one - thereby magically increasing pixel density by a third as cynics would argue. In the end that's more arguing about nomenclature than anything else though. – Voo Jul 22 '16 at 23:05
44

Pixels per inch is essentially the digital equivalent of dots per inch - it's an arbitrary measurement that lets you know how sharp an image is. A pixel itself isn't 'standard' on a modern LCD screens - there's different arrangements of 'subpixels' that make up colours, like rgb LCDs, which consist of equal red, green and blue sub pixels; rgbw, which adds a white subpixel to that; or pentile.

For reference, here's an rgb layout

enter image description here

Each 'trio' of red, green and blue is a pixel, and the pixels are vaguely rectangular here.

Compare that to pentile.

Here, each pixel is one blue, 2 reds and 2 greens - it's not even a quadrilateral! By Matthew Rollings at English Wikipedia, CC BY 3.0, https://commons.wikimedia.org/w/index.php?curid=14914205

A third point of reference would be black and white screens - like e-ink or high resolution medical displays. Here, each pixel would be a single colour, black.

As such, there's no such thing as a standard pixel, or a standard size for said pixels. PPI is a measure of how small pixels in terms of pixels per square inch but these pixels need not be a standard size, or even shape or arrangement. PPI also doesn't really mean anything unless it's the same display layout. An rgb screen would have higher effective resolution/sharpness than a pentile screen in many situations.

Personally, I consider pixel pitch a complimentary measure since it also takes into account the spaces between subpixels and the fact that there are subpixels. Gaps between pixels affect the perceived quality of a display a fair bit. Two displays with the same DPI and different pixel pitch will have very different qualities.

enter image description here

tldr: There's nothing standard about the size or shape of a pixel or even the elements that make up a pixel.

  • 7
    It so trippy squinting your eyes and watching the first image turn white with colored letters. +1\ – ApproachingDarknessFish Jul 24 '16 at 21:52
  • I think the shadow mask picture you included is a bit confusing because those are different ways of constructing a CRT screen, and there's no 1:1 correspondence between phosphor dots on a CRT screen and displayed pixels. The distance shown by the white arrow would be described as the dot pitch of the screen, and you would want it to be a bit smaller than the distance between displayed pixels. – nekomatic Jul 25 '16 at 11:51
  • At this point that first RGB layout is more myth than reality. While the positional ordering is still true, it's more like comparing telescopic binary stars (pixels) to spectroscopic binary stars (individual LCD elements) than what that diagram shows on every monitor I have examined with a glass. I got here trying to figure out why cleartype always fails with color fringes no matter what settings I try so it's based in reality not an educated guess. – Joshua Jul 25 '16 at 15:43
20

The other answers are missing the point here. To quote from Wikipedia,

In digital imaging, a pixel, pel, dots or picture element is a physical point in a raster image, or the smallest addressable element in an all points addressable display device.

There is no a priori restriction on the physical size of a pixel. I've used devices with 500 µm2 pixels; you could justifiably claim that a 5-mm diameter photodiode is a single-pixel device.
There's a "sensible" lower limit on pixel size which is on the order of a wavelength, only because it's physically impossible to produce an image of higher resolving power than that. Even there, you could imagine a ultra-high-energy gamma-ray sensor whose pixels are a few femtometers in size.

  • 3
    @Mooing I think the point is that it's a practical, not physical, limit. If the pixel size is smaller than about one wavelength, diffraction will make the image from the individual pixel spread out and overlap adjacent pixels. There's nothing conceptually preventing a single pixel from outputting several wavelengths simultaneously (pixels on most real devices probably do). An earbud is essentially a single "sound pixel" that outputs multiple wavelengths at once. Given an array of earbuds outputting different wavelengths, diffraction prevents distinguishing the sound from individual earbuds. – j_foster Jul 22 '16 at 23:24
  • 3
    You mean LED, right? Photodiode is quite an opposite device, it does not emit light. – n0rd Jul 23 '16 at 1:45
  • 2
    @n0rd: Cameras have pixels too. A thing can have an imaging resolution without being a thing that emits light. – Lightness Races in Orbit Jul 23 '16 at 12:59
  • 4
    @mooingDuck: first of all, those earbuds are putting out longitudinal, not transverse, waves, so the bud size is irrelevant. Second, Sound production is not a quantum operation. Photon absorption is a quantum operation, so you can't go by wavelength alone. – Carl Witthoft Jul 23 '16 at 13:16
  • 1
    This is the correct answer a pixel is a sample and there is nothing that says how large your element needs to be. In fact some old stadium displays had lightbulb sized pixels. – joojaa Jul 23 '16 at 14:52
13

In this case [*] is the inch that has a fixed size of 2,54 cm. The dimension of the pixels can vary.
The PPI is the number of pixel that you can count in one inch of your device.

enter image description here
I think that sometimes a picture is more effective of 100 words. Read more from the source of this figure.

In the picture above the left one has a number of PPI less (10) then the right one (20).
More it is high the number of PPI, more realistic it will appear the image you are looking at.


[*] O.T. Trivial note: it should be fixed at least after the last modification of the 1983. Nowadays the inch is the unit of length in the Imperial and in the United States customary systems of measurement. In past the inch was used by many different systems. One of the earliest legal definition of the inch was set out in a statute of Edward II of England in the 1324, it was used before and it was fixed only in the 1959 when the yard (36 inches) was fixed to to 0.9144 metres. The meter instead belong to the International System of Units, it was originally defined in 1793 thanks to Louis XVI, and it was (1983) correlated to the speed of the light in vacuum. The speed of the light in vacuum is the real physical constant, but until your mobile will not be able to travel at relativistic speed, you will not notice. :-)

  • 0.9144 metres may be a yard (36 inches) but it sure ain't an inch. In several languages, the word for inch is the same as for thumb, which was the lay definition of an inch (the width of a man's thumb). – Adrien Jul 23 '16 at 4:05
  • The meter is now based on speed of light, but in 1793 it was based on the size of the Earth, nominally one quadrant of the meridian from the equator to the North pole through Paris. – dave_thompson_085 Jul 23 '16 at 9:24
  • Despite having signed the preliminary decrees predating the metre definition, Louis XIV was likely instead supporting the old system using the six feet toise. In any case he was beheaded before the new metric system mètre étalon was built, equal to the ten-millionth of the distance from the equator to the North Pole. – jlliagre Jul 23 '16 at 9:36
  • @Adrien thanks for the spot. Yesterday, continuing to correct to reduce the length, I ate the important part... better to go to sleep earlier sometimes ;-) To Dave, of course the Speed of the Light was postulated by Einstein centuries after, in the 1905, in natural unit it is 1. To jilliagre, right and proper, but it was over the extent of an OT note to write all the information about it... I find interesting the link between inches and this case, and the link between kings and science... – Hastur Jul 23 '16 at 9:38
8

There are two different but linked concepts of pixel:

The classic (physical) one, which is the basic unit of a display, the smallest dot that a screen can turn off and on, generally composed by three colors (RGB).

But as it vary a lot from one device to other, when you want to show anything of a certain size in different devices ti becomes complicated: 500 pixels in a desktop are about 15cm, but only 3cm in a smartphone. So for web developers there is another concept: 1px = 1/96th of one inch, so if you want to create a button of half an inch, you define its width as 48px and delegate the "real" size calculations to the browser. But even in this case it is just a rough approximation, because the given proportion between both concepts (named devicePixelRatio in browsers) is usually just an integer value and it depends on the system giving the right value to the browser.

For instance: the iPhone 6/7 Plus has a display FullHD (1080 x 1920 "real" pixels), but for a web developer it has 414 x 736 "logical" pixels)

  • The browser might not know the size of the screen. So there is no way to reliably render something to be a specified number of cm in size. – kasperd Jul 22 '16 at 17:33
  • 2
    The 1px = 1/96 inch is true when the output is for the printed page. For screens, it's more complicated, and it has to do with the device resolution and the typical distance from the eye to the screen. But the idea was that a px would subtend the same angle of your field of vision as a traditional pixel would on a typical monitor (often 96 pixels per inch) at a typical viewing distance. – Adrian McCarthy Jul 22 '16 at 18:32
6

No.

Way back in the (multisync CRT) day, we had monitors of the same size and resolution with different dot pitches (pixel sizes). Dot pitches (sizes) were sometimes so large that they overlapped and made for slightly fuzzy looking pictures, so smaller pixels (even at the same resolution) were generally preferred.

3

Yes, the size of a pixel is 0x0. Note that this definition works for all units of distance: 0cm x 0cm, 0 inches x 0 inches, 0 miles x 0 miles, etc.

I assume you're asking whether there is a standard separation between pixels. The other answers deal with that question quite well, so I'll use this space to belabour my pedantic point instead ;)

A pixel is a point sample; it specifies the colour value at that point; the surrounding points (those between pixels) have no associated colour value. If we want to assign a colour value to points in between pixels (because, say, our device has several display elements between each pixel), we must interpolate the surrounding pixels somehow.

One very crude method is the "nearest neighbour" approach. This creates a Voronoi diagram by taking the value of each intermediate point to be equal to that of its nearest pixel. This ignores all gradient information, resulting in jagged, aliased displays. Unfortunately it's also a very common method of interpolation; so widespread that it has lead to a common confusion/conflation between pixels (which can't be seen, as they have no size) and the Voronoi cells rendered around them (which can be seen).

The results of various interpolation methods, including nearest neighbour, can be seen here. That page discusses "scaling up an image to 4x the area"; that's equivalent to using a display with a display unit density 4x higher than the pixel density; or, more generally, resampling the image.

For a friendly explanation, see http://graphics.cs.wisc.edu/WP/cs559-fall2014/2014/08/29/what-is-a-pixel-and-what-is-a-point-sample

For a less friendly explanation, see http://alvyray.com/Memos/CG/Microsoft/6_pixel.pdf

  • whilst we may refer to a point, it's not a point of infinitesimal width. 0 x 0 is invisible, and not useful for any kind of display not even theoretically. Pixels in real and virtual devices have non-zero width and height – Adrien Jul 25 '16 at 10:20
  • @Adrien you are wrong. In signals theory they are infinitesimal in size. Their emittance behavior is that of a 2D Dirac distribution (infinitesimal size, infinite value, finite integral). Of course real devices need some surface are to create visible light, but that doesn't change the fact that a theoretical pixel is a point sample and the real device is an approximation of that point sample filtered with an ideal lowpass filter. – Joe Jul 25 '16 at 12:09
  • well it's a bit of a leap to go from that to claim that pixels have 0 width and height. They don't in reality, and I don't really buy the theory either when it comes to pixels. Yes for points, but pixels aren't necessarily points (singularities). – Adrien Jul 25 '16 at 12:15
  • @Adrien Pixels are not physical objects, in the same way that "42" isn't a physical object. I used the term "display units" to refer to things like LEDs, spots of ink, etc. since they are not pixels; although the input signal driving them can be specified using pixels. Likewise, the devices at the other end (CCD elements, rod/cone cells, photomultiplier tubes, etc.) aren't pixels either, but their output signal can be represented as pixels. – Warbo Jul 25 '16 at 12:25
  • I guess then that the term "pixel" is widely abused. When you zoom in on an image using a paint app, until you see the big coloured squares, are they not pixels? Those are squares, not points. – Adrien Jul 25 '16 at 12:27
3

The other answers have already covered the fact that there might be different sizes of pixels.

However, the question was (emphasis mine):

So does Pixel have a standard size?

There is, in a way, a "standard" of 96 pixels per inch in Windows. This is what led to the creation of DIP which "assumes" 96 DPI (as mentioned in the link).

protected by Community Jul 5 '17 at 2:30

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged or ask your own question.