The Myth of DPI

[ken-w]

Hi all - over the last few months I've seen a number of posts where people get hung up on equating the dots per inch (dpi) of a photo to digital image quality ("my camera only shoots at 72 dpi, Bill's shoots at 150 dpi - Bill's photos must be higher quality than mine").

In an effort to get the word out that DPI has absolutely nothing to do with digital image quality, I've posted a little web page on the subject: www.rideau-info.com/genealogy/digital/dpi.html

Constructive critisism is always welcome

[Jorge]

Hi Ken

I do understand what you getting at but I think your eager to prove your point has brought in some false arguments.

You’ve got two similar images in your website that’s labelled 1000 DPI and 10 DPI respectively – and hey, who can see the difference? Point proven! DPI doesn’t affect image quality!

But as you say later in the page neither of these pictures are actually what you write. Firstly DPI as a measure does not apply to a computer monitor – here PPI (pixels per inch) is the relevant measure. Secondly the value of PPI depends on the resolution and size of the monitor. Viewing your website on a 15 inch monitor showing 1024 * 768 pixels will give the approximate value of some 70 PPI for BOTH pictures. So no point proven here – only confusion is added.

But like I said I know what you mean. DPI is not a relevant measure for digital resolution, but it is relevant for prints. Prints in magazines are somewhere between 300 and 400 DPI, while newspaper quality is more like a 150 DPI. To take full advantage of a given DPI you’ll have to have a corresponding amount of data in you’re digital file. Since every print these days derives from a digital file it’s important to establish the link between pixel data and the output DPI.

As far as I know (I’m no expert in printing) you’ll need one pixel for one dot. To print a 5 x 5 inch at 300 DPI will take 5x300x5x300 pixels or 2.250.000 pixels. It’s not that mysterious. These pixels will have to have a colour dept of 32 bits to correspond a normal four-colour (CMYK) printing technique. This means that a standard 2 mega-pixels camera can’t do the job, since each pixel don’t have the sufficient colour dept. As far as I know the colour dept is more like 8 bits pr. pixel, and representing only the value of R, G or B, but that’s another story.

[Pekka]

But as you say later in the page neither of these pictures are actually what you write. Firstly DPI as a measure does not apply to a computer monitor – here PPI (pixels per inch) is the relevant measure. Secondly the value of PPI depends on the resolution and size of the monitor. Viewing your website on a 15 inch monitor showing 1024 * 768 pixels will give the approximate value of some 70 PPI for BOTH pictures. So no point proven here – only confusion is added.

PPI or DPI - on a computer monitor they are practically the same. Dot = Pixel in that context.

DPI of the monitor changes between different resolutions. Source DPI (image DPI) is transparently mapped by operating system to target DPI (monitor), pixel in source = pixel on target. This means that source DPI changes when target resolution changes, so source DPI has no real meaning on a computer screen, unless viewer application takes it into account and remaps the pixels (in Photoshop: view print size).

As far as I know (I’m no expert in printing) you’ll need one pixel for one dot. To print a 5 x 5 inch at 300 DPI will take 5x300x5x300 pixels or 2.250.000 pixels. It’s not that mysterious. These pixels will have to have a colour dept of 32 bits to correspond a normal four-colour (CMYK) printing technique. This means that a standard 2 mega-pixels camera can’t do the job, since each pixel don’t have the sufficient colour dept. As far as I know the colour dept is more like 8 bits pr. pixel, and representing only the value of R, G or B, but that’s another story.


Has anyone measure the real RGB dpi of e.g. Epson 2200?

Only RGB printers like sublimation printers can print pixels, without a raster. Normal inkjets have only max 7 prime colors and they can produce full gamut only by rasterizing and overlapping clusters of dots to fool eye to see certain colors. I doubt there is a fixed DPI value in the result.

[ken-w]

Jorge wrote:
But like I said I know what you mean. DPI is not a relevant measure for digital resolution, but it is relevant for prints.

Hi Jorge - thanks for your comments. I agree with you to a point - but again we're getting mixed up a bit. My main point is that from a digital photo perspective, DPI is irrelevant, no matter how you set the DPI for your image, the image stays the same (pixels are not remapped). The myth that seems to be out there (from several posts on this on other digital photo forums) is that DPI is somehow a measure of the quality of the original digital image - it isn't.

The two images on my website are as you note, most likely being displayed on either a 72 ppi or 96 ppi monitor display. But each photo has a much different internal dpi - the point is that you can't tell the difference (since there is no difference) - if you download both photos you'll see the first as a 1000 dpi photo and the second as a 10 dpi photo and they will look the same no matter what display device you use, including being printed if you print them at the same physical size (without altering the dpi settings).


As far as I know (I’m no expert in printing) you’ll need one pixel for one dot. To print a 5 x 5 inch at 300 DPI will take 5x300x5x300 pixels or 2.250.000 pixels.


Actually it's not quite that simple as Pekka has pointed out. Many printers do remap the photo and most inkjet printers use a blended dot. I do some desk top publishing, proofing usually on a 300 or 600 dpi laser and then printing using a commecial digital printer (at very high dpi). The photos all have the same internal dpi setting (usually just the default 72) but they do look much better from the commercial digital printer because of its higher DPI.

On your final note - colour depth is a whole other thing - we won't go there in this post :D