DPI - Unravelling A Mystery

[RocklandPaul]

There is actually a pretty good, accurate, and comprehensive explanation in Wikipedia under "Pixel":


http://en.wikipedia.org/wiki/Pixel


The whole article is worth reading, and re-reading.


Here is a relevant sub-section copied from that Wikipedia "Pixel" entry:





<span>A pixel is generallythought of as the smallest single component of a digital image. The definition is highly context-sensitive. For example, there can be "printed pixels" in a page, or pixels carried by electronic signals, or represented by digital values, or pixels on a display device, or pixels in a digital camera (photosensor elements). This list is not exhaustive, and depending on context, there are several terms that are synonymous in particular contexts, such as pel, sample, byte, bit, dot, spot, etc. The term "pixels" can be used in the abstract, or as a unit of measure, in particular when using pixels as a measure of resolution, such as: 2400 pixels per inch, 640 pixels per line, or spaced 10 pixels apart.


The measures[url="http://en.wikipedia.org/wiki/Dots_per_inch][i]dots per inch[/i][/url][i](dpi) and[/i][url="http://en.wikipedia.org/wiki/Pixels_per_inch][i]pixels per inch[/i][/url][i](ppi) are sometimes used interchangeably, but have distinct meanings, especially for printer devices, where dpi is a measure of the printer's density of dot (e.g. ink droplet) placement.[/i]<sup class="reference" id="cite_ref-6"][url="http://en.wikipedia.org/wiki/Pixel#cite_note-6]<span>[i][[/i][i]7[/i]<span>[i]][/i][/url]</sup>[i]For example, a high-quality photographic image may be printed with 600 ppi on a 1200 dpi inkjet printer.[/i]<sup class="reference" id="cite_ref-7"][url="http://en.wikipedia.org/wiki/Pixel#cite_note-7]<span>[i][[/i][i]8[/i]<span>[i]][/i][/url]</sup>[i]Even higher dpi numbers, such as the 4800 dpi quoted by printer manufacturers since 2002, do not mean much in terms of achievable resolution.[/i]<sup class="reference" id="cite_ref-8"]<span>[9<span>]</sup>


The more pixels used to represent an image, the closer the result can resemble the original. The number of pixels in an image is sometimes called theresolution, though resolution has a more specific definition. Pixel counts can be expressed as a single number, as in a "three-megapixel"digital camera, which has a nominal three million pixels, or as a pair of numbers, as in a "640 by 480 display", which has 640 pixels from side to side and 480 from top to bottom (as in aVGAdisplay), and therefore has a total number of 640 &times; 480 = 307,200 pixels or 0.3 megapixels.


The pixels, or color samples, that form a digitized image (such as aJPEGfile used on a web page) may or may not be in one-to-onecorrespondencewith screen pixels, depending on how a computer displays an image.


I think that one place where a lot of people get confused is that they forget (or don't realize) that neither "Dots Per Inch" nor "Pixels Per Inch" says anything specifically about the physical size of that "Dot" or "Pixel". All it is saying is how many of these (whatever size) dots or pixels there are in one inch. Many people leap to the conclusion that if a printed image is at "600 DPI", that each Dot must therefore be 1/600 inch in diameter. Not necessarily true!


There's a lot more to this discussion, but start by reading the Wikipedia entry.


-Rockland Paul

[Keith B]

Photoshop renders pixels the same regardless.


Pixels from my 5D Mk2 were .004" at 240dpi


Pixels from my 40D were .004" at 240dpi


If I set the image size to 300dpi in Photoshop (resample off) both become .003"s.


72 dpi = .014"


So pixel size is determined when you give it the arbitrary value in Image editor. But ultimately determined by the size you print it.


Rules of thumb are 304.8dpi is the standard for (magazine-print ad) printing, set in Europe some time ago. It is actually a perfect 12 pixels/mm.


200dpi is the cutoff of where the human eye can begin to depict the pixels instead of smooth image.


Canon Raws are 240dpi by default (happy medium?). Jpegs 72dpi.


In practical terms, when you print an image without interpolation the pixel will scale down to what it needs to be. If you scale it up it will scale up and once it goes beyond the 200dpi point the pixels will start to become apparent.


So I can print an image from my 5D Mk2 at 28 inches wide before I detect the pixels (200dpi).


If you scale it down, once you get below 75% or 17.55" it will start to throw out pixels (down sample), reducing image quality.





In summation, in the camera pixels can be different sizes but once in an image editor pixels become a standard size.DPI determines pixel size.


You can have too many pixels. Scaling down can actually reduce image quality.





I hope this helps.

[Citronsyra]

Interesting it is I must say!





But Keith: How did you came up with the fact that the maximum print of an image is 28 inches wide for your 5D MkII?

[Keith B]

at 28 inches wide my images will be at the 200dpi point and you will start to see blockiness after that. I can obviously go as large as I want but the pixels will start to become apparent.


Although once an image gets so large, one has to be a good distance back from it to view the whole image, then the eye is more forgiving.


Like billboards, that you see on the roadside, are usually printed at 100dpi and they will have no problem interpolating those images either because the eye isn't going to detect all the fine details from 100 feet away.


Photoshop does a really good job of resampling, If your image starts out sharp you can easily double the size with out loosing much detail or at least be able to see it with the naked eye at a reasonable viewing distance.

[RDKirk]

Photoshop does a really good job of resampling, If your image starts out sharp you can easily double the size with out loosing much detail or at least be able to see it with the naked eye at a reasonable viewing distance.

Or better than double it, depending on whether the original image captured sufficient detail for your purposes.


For the portrait work I do, an important pictoral factor is that viewers have a distinct expectation of sharpness: A portrait is considered "sharp" if the facial hair is sharp when the face is seen at a size at which they'd expect to resolve facial hair with the naked eye.


In practice, if the head size is about 3 inches or larger on the final display (an apparent "distance" at which the naked eye would resolve a person's hair), the hair must be resolved for the image to be considered "sharp." (I always assume viewing distance to be "reading distance" because I find that when viewers are engaged by a photograph--my intent, of course--they will move as close as physically possible to see the detail they expect in a photograph. They don't care about theoretical viewing distance rules!) If the head size is less than 3 inches in the enlargement, viewers don't mind blurred hair because the naked eye doesn't resolve hair when a person is seen from such an apparent distance.


But viewers of portraits don't care to see skin flakes or hair mites, no matter how great the enlargement or how close they get--if the hair is sharp, that is sufficient. The fortunate thing for digital photographers is that hair is easy to interpolate in software--it's just a tone without internal detail. As long as the facial hair is resolved in the original capture, the image can be successfully upsampled to nearly any size.


So the deciding factor becomes the scale of the face in the image, and the photographer has to work within those limits. Any DSLR can resolve facial hair in a head-and-shoulder portrait, thus even the 4- to 6-megapixel older DSLRs did well enough for those kinds of portraits. An 8- to 10-pixel DSLR can resolve hair up to around a half-length portrait, and a 12-megapixel DSLR can get to 3/4-length. At 21-megapixels, hair is resolved in a full-length portrait. This is presuming the head size is enlarged in the final display to 3-inches or greater.


Is it possible to create a successful portrait even if the hair is not resolved in the original image? Yes, but the photographer has to work around that limitation--printing on canvas, for instance, or using Corel paint or a soft-focus treatment to make the lack of detail "intentional."


A landscape, OTOH, is a totally different situation. Viewers expect a landscape image to reveal more and more detail the greater its enlarged and the closer they get to it. Viewers become frustrated at the point they see that there is no more detail to be revealed.


Thus, the resolution requirement for a landscape camera is practically infinite. The photographer has to work within the limits of the equpment by limiting the size of the final display (little or no upsampling) or using as high a resolution camera as is available.

[peety3]

As I (slowly) learn the realities of this, I've come to the following rough conclusions:


Good prints look best if they're printable at 300ppi. Large prints may be OK at 150-200ppi, on the assumption that no one is standing right at the print and staring at that level of detail.


Printers fit pixels to physical dimensions (so a 2400x3000 pixel image file printed at 8x10 is sized for 300ppi) and then those pixels are mathematically expanded to suit the dpi of the printer head (so a print head that prints at 1200x2400dpi would convert those pixels above into a 4x8 rectangle of dots per pixel).


With those conclusions, my life has become easier.