: PPI vs. DPI confusion (not a duplicate! NOT the same question that's been asked!) I was also confused about the meanings of PPI vs. DPI and so I read a few tutorials on Google (including

@Jamie315

There are multiple definitions of a pixel and this is unfortunate.

The problem with the physical LCD pixel descriptions you provided is that they are trivially wrong (the vast majority of displays use multiple light elements to produce a single "pixel"); they are involved primarily with specific implementations (of which there are many and various); and they are not relevant to the majority of producers of image content (especially those who are actually confused about the distinctions).

Regardless, a physical LCD pixel has a dimension since it is a physical object that can be measured with a ruler. This is one definition of a pixel. But this is the wrong definition for image editing and manipulation.

The problem with DPI or printing definitions is again, an implementation problem. For printing there is a tighter target, but commercial printing can be stochastic or FM screening, and the size of the screens vary. In addition each color screen is sampled at a different angle, so the screens themselves to not have a direct correlation to the apparent grid-like nature of the dominant raster-image storage methods.

Again, these are physical dots and they are not one-to-one representations of pixels stored in an image file. Obviously, this is not a definition of a pixel (yes they are picture elements, but the etymology of the word only tells us where it came from, not the specific meaning).

The thing is: the pixels people are concerning themselves with are the ones we manipulate in e.g. photoshop or GIMP etc.

These pixels have no size. They are dimensionless numbers. To ask what size they are is to misunderstand what they are. To ask what color they are is a misunderstanding. When you ask someone how big they want the image and they say "300ppi" they might as well say "purple monkey dishwasher". If someone wnats to know the size of a raster image pixel, the only answer (aside from a blank stare) is mu.

When we talk about resizing an image, we usually mean resample: add or remove data. This is where image quality degradation comes from. If we are not actually re-sampling a physical object with a higher density of samples, then we are making up data.

When we say there are color channels, we forget that using the data for color information is a convention, not a requirement. The red channel might be depth data, the blue channel might be brightness, the green channel might be time of day in seconds.

When we make an image "300ppi for print" what we are doing is: deciding on a target physical size on the paper we want for the finished item; considering the expected line screen halftone process being used by our supplier; providing twice as much data so that the when the plates are made, the resampling process can accurately reproduce the original signal. We can usually assume that the print providers have equipment designed around 100-133 line screen, and we like to give a little more data, so the rule of thumb works and we don't need to calculate.

We usually get pretty good results with this rule so it gets spread around, and it works. But the key here is to focus on the data, not the PPI. The PPI is mutable and derived and has very little value without a physical target.

NASA has an inverse relationship with with imagery. Ignoring that they use different intermediary image storage methods etc, the products they put out to the public are using typical 3-channel raster images and often have a resolution quoted with them. Only they quote in "meters per pixel." They can do this because, again, the pixels are dimensionless point samples that are fixed at the time of creation. The resolution of "3 meters per pixel" is the threshold below which the depicted objects are smaller than 2 point samples and therefore suffer from aliasing

For more reading on 2x data sampling ( Shannon-Nyquist Sampling Theorem ). Note that this is the reason "CD quality" is 44kHz: the human ear has about a 21kHz range and so music files sample at double that in order to properly reproduce the original signal without noticeable aliasing.

For more reading about pixels as point samples see (A Pixel is not a Square)

Vote Up Vote Down

@Kimberly620

PPI is just a conversion factor. It is essentially same thing as DPI. The primary difference between DPI and PPI is that PPI describes image elements that can have more than one color per channel (printers dots are either there or not there).

PPI and DPI are ONLY meaningful if you manufacture/buy displays or prints. Since prints are always manufactured its allways somewhat relevant. But if you are using monitors, other than using one specific monitor, then the value is almost never relevant!

The problem is that people want PPI to be more fundamental than it is.

Notes:

There is no conflict between PPI being a feature of the physical device and it being a metadata in the image file. The other qualifies the device the other is a indication what the designer was wishing when creating the image. The other is a statement of fact and the other is a statement nore like: " I designed this picture using a ruler unit of X so that the image is 20cm wide". This makes sense from a manufacturing point of view, ie print, or if you happen to specify what monitor to buy. But a mobile phone user or a projector user really has no way to comply with your wishes in many cases. So they will just ignore you as stupid.

Adobe in their infinite wisdom never considered that a image file could actually have "no size", forcing you to think about this even if its irrelevant to you. Image files quite often have no specifier for the PPI or DPI value. 1 Pixel being different size on different devices, but that is ok and how it should be do.e in many cases.

Vote Up Vote Down