Image size and ppi

[enduser]

A customer sends me a jpeg image from a Pentax DSLR and in Photoshop the 3mb, 6 mpixel  image is now said to have a file size of 17mb.  I don't understand that.  The second thing I don't get is that Photoshop says the image is 41 inches x 27 inches and has a resolution of 72 pixels per inch.  First, does the image have a specific "size" at this point?  Isn't it just a collection of readings on a 2000 x 3000 array of sensors?  Second, why only 72 ppi?  My early 300D Canon gives me 180 ppi jpeg images. (And 240ppi RAW images)

Sorry about the messy way I've asked, and thanks in advance for any help.  (Just point to a good reference will do)

[Schewe]

JPEG is a compress file format so it's normal for the image to be larger after opening (and decompressing) the image. Also not that the camera can usually control what the PPI setting is set to. If not set then it's common to default to 72PPI (in essence, no specified resolution). You can change the Image Size in the Image Size delog. As long as you uncheck the Resample open all it will do is change the dimension/PPI resolution ratio–it won't actually resample the image. As long as you don't save over the JPEG (save as TIFF or PSD) then there will be no alteration of the original file.

[mfryd]

A customer sends me a jpeg image from a Pentax DSLR and in Photoshop the 3mb, 6 mpixel  image is now said to have a file size of 17mb.  I don't understand that.  The second thing I don't get is that Photoshop says the image is 41 inches x 27 inches and has a resolution of 72 pixels per inch.  First, does the image have a specific "size" at this point?  Isn't it just a collection of readings on a 2000 x 3000 array of sensors?  Second, why only 72 ppi?  My early 300D Canon gives me 180 ppi jpeg images. (And 240ppi RAW images)

Sorry about the messy way I've asked, and thanks in advance for any help.  (Just point to a good reference will do)

Photoshop is telling you that the uncompressed image is about 18 megabytes.  Each pixel is 3 bytes (one red, one green, one blue), so you should expect about 18 megabytes of uncompressed data for a 6 megapixel image.

Photoshop is also telling you that if you printed the image at 72dpi, then the image would be 41 by 27 inches.  If you change the dpi to 300 dpi (without resampling), it will tell you the image is closer to 9.9 by 6.5 inches.

Your image doesn't really have a specific size other than the dpi field, which gives a suggestion as to how it should be printed.  Yes, the image really is just a 2,000 by 3,000 array of pixels.

Lots of software defaults to suggesting 72 dpi, as this is a traditional resolution for computer monitors, and 1/72nd of an inch is about equal to one "printer's point" - a standard unit of measurement in the printing industry.

If your Canon 300D is giving you 180 ppi JPEGs and 240ppi RAW files then this is an artifact of the preference set for your camera and RAW software.

[neile]

Lots of good explanation already posted, but just to clarify something in mfryd's post: all the "dpi"s in that post should really be "ppi"s. DPI is a measure of how many little droplets of ink your printer spits out onto the page, and has no relation to the number of pixels you have in your image. PPI is the correct term to use when you are looking at pixels on screen and are trying to figure out how big the print will be.

This is reflected in the Photoshop Image Size dialog where the resolution field is "pixels/inch", not "dots/inch".

Neil

[enduser]

Thanks all.  Amazing how straightforward it is once it's explained.  Thanks again.

[mfryd]

Lots of good explanation already posted, but just to clarify something in mfryd's post: all the "dpi"s in that post should really be "ppi"s. DPI is a measure of how many little droplets of ink your printer spits out onto the page, and has no relation to the number of pixels you have in your image. PPI is the correct term to use when you are looking at pixels on screen and are trying to figure out how big the print will be.

This is reflected in the Photoshop Image Size dialog where the resolution field is "pixels/inch", not "dots/inch".

Neil

It's an old habit.   I remember Linotype imagesetters that had a resolution of 1,200 dots per inch.  When printing images, we could set the halftone screen to whatever we wanted.  We still called it 1200 dpi (the resolution of the printer) even when printing an image with a 100 dpi halftone screen.

With inkjet printers, I get confused.  The printer itself may be able to place ink dots with 1/1440 of an inch accuracy, but it can also vary the amount of ink (hence size) of that dot.   Not only that, but few inkjets use a traditional halftone screen with a regular halftone dot pattern.  Many printers use proprietary (and undocumented) dithering patterns where there might not be anything corresponding to a traditional halftone dot.

 

[teddillard]

Yes, the DPI/PPI thing is one of those issues that seem designed to confuse people.  I go on a regular rant on it in every class I teach, and I'm not sure but I think there's a rant in all of my books.    

Here's the thing.  There's nobody (now, anyway) who actually uses "dots per inch".  In Photoshop you're using pixels per inch, and pixels (in Photoshop anyway) are little squares of color.  In printers if you must use "DPI" then you're talking "droplets per inch", not "dots", and the droplets are really better measured by their picoliter size than their, well, density, since they vary.  Unfortunately the marketing guys feel they can sell more printers by using DPI incorrectly. 

In the simplest terms, you're converting little squares of color created by mixing light to droplets of ink...  two entirely different ways of working with resolution and color.  This is what "dithering" is, basically- translating those little squares of light into drops of ink so the color looks right.

The offset printing industry actually is the only industry that does use a practical application of "DPI" to describe the screen resolution of a separation, and they, ironically, refer to the screen required to make it- for example, a 165 line screen- or "LPI"...  so the one industry that uses DPI calls it "LPI".  Go figure.  

In any case, using "DPI" in any way for digital photography is simply incorrect.  (And if Apple is still doing it in Aperture, they should be slapped.  )

[Luca Ragogna]

The offset printing industry actually is the only industry that does use a practical application of "DPI" to describe the screen resolution of a separation, and they, ironically, refer to the screen required to make it- for example, a 165 line screen- or "LPI"...  so the one industry that uses DPI calls it "LPI".  Go figure.  

Just to be a total geek about this.

LPI and DPI are 2 different things again. LPI refers to an image after it's been screened and DPI is an unscreened resolution. Our CTP is 3600 DPI but we use it to output plates at 200 LPI. We can change the screening to 100 or 150 LPI but they will always be imaged at 3600 DPI.

Also laser printers use DPI too (and pretty much anything that uses a laser to create an image). The image is created with dots of a set size a certain distance apart.

[mfryd]

Just to be a total geek about this.

LPI and DPI are 2 different things again. LPI refers to an image after it's been screened and DPI is an unscreened resolution. Our CTP is 3600 DPI but we use it to output plates at 200 LPI. We can change the screening to 100 or 150 LPI but they will always be imaged at 3600 DPI.

Also laser printers use DPI too (and pretty much anything that uses a laser to create an image). The image is created with dots of a set size a certain distance apart.

This was my understanding, and leads me to believe that dpi or ppi are not really applicable concepts to modern inkjet printers.

Modern printers use variable size dots with very precise placements.  On an Epson 3880 the dots can be positioned with an accuracy of 1/1440 of an inch on one direction and 1/2880 of an inch in the other.  The positioning grid is not square, and the dots overlap.

At best, we might describe this as a 1440 or a 2880 dpi printer.   There is nothing about the hardware that would make this a 360 dpi photo printer.

If this was a traditional PostScript typesetter, we could talk about the halftone screen frequency, for instance we might be printing with a 100 or 150 line per inch halftone screen.  Once we start talking about half tone screens we need to accept that the orientation of the halftone screen might not be aligned with the pixels in our image.  Thus if we are using a 100 line per inch halftone screen, we want to send the imagesetter a 200 pixel per inch image.  This ensures that the pixels resolution is much higher than the halftone screen resolution, minimizing degradation from any "re-sampling".  If our image has high contrast details (i.e. a line drawing), we benefit from sending an even higher resolution image as PostScript uses "partial dotting" in creating the halftone screen.   "Partial Dotting" maximizes the preservation of details that are higher frequency than the halftone screen.  Thus we are almost always get better quality by sending a  higher resolution image as opposed to downsizing to "match" the halftone screen.

With modern inkjet printers, the situation becomes cloudy.  The algorithms Epson uses to go from the pixels in the image to the dots on the paper are a secret.  We just don't know what the algorithm does, or exactly how it works.

The common belief is that algorithm uses a square grid with 360 (or 720) "cells" per inch, and that this is the best resolution to use when printing.  Obviously Epson is not using a traditional halftone screen.  They use sophisticated dithering to make tints of the native inks.  It is not at all clear if the dithering works on each "cell" in isolation, or if it takes into account adjacent pixels.  It is also not clear if there is anything similar to PostScript's "partial dotting" which allows the printer to take into account the additional information available when image pixels are smaller than 1/360 of an inch.

Without this information we are all just guessing.

Can someone explain the rationale behind calling an Epson 3880 (which places dots on the page with an accuracy of at least 1/1440 inch) a 360 dpi (or ppi) printer?

[Bart_van_der_Wolf]

Can someone explain the rationale behind calling an Epson 3880 (which places dots on the page with an accuracy of at least 1/1440 inch) a 360 dpi (or ppi) printer?

360 dpi doesn't make sense, 360 ppi is possible, but the native resolution (as reported by the printerdriver) of the printer is 720 ppi with an appropriate selection of parameters and media. It's the smallest physical (RGB!) pixel size that can be printed (and resolved, e.g. in cases of vernier acuity).

Cheers,
Bart

[PeterAit]

The image size, in inches, and the resolution, in PPI, are totally arbitrary. PS assigns default values when you open an image that does not include these data. In this case, the arbitrary 72 ppi leads to the stated image size. You can change it to anything you want using the Image Size command, with resampling turned off.

[Ernst Dinkla]

It's an old habit.   I remember Linotype imagesetters that had a resolution of 1,200 dots per inch.  When printing images, we could set the halftone screen to whatever we wanted.  We still called it 1200 dpi (the resolution of the printer) even when printing an image with a 100 dpi halftone screen.

Your offset press doesn't know a native dpi or lpi, in theory it can print say a 250 LPI raster next to an image that had a 150 LPI raster. The (photo) imagesetter that creates the halftone film or plate can have a 1200 LPI resolution (usually higher though) which means its laser can put 1200 lines with 1200 equal sized spaces in between. With that laser it can build all kinds of shapes (fonts) and patterns on lith film or offset plate, among them halftone screens. Up to 120 LPI halftone raster in this case as there is a rough rule that the imagesetter has to have about 10x the resolution of the halftone screen it has to create. It is unusual to qualify the offset printwork by the resolution of the imagesetter, a printshop will quote its maximum halftone screening if it specifies quality.  LPI is somewhat confusing too though, halftone screens usually have elliptical dots or round dots, the last can get a square shape at 50%. There are line screens, wave screens, whatever way you can split paper white and ink black at 50% and let it gradate either way.

With inkjet printers, I get confused.  The printer itself may be able to place ink dots with 1/1440 of an inch accuracy, but it can also vary the amount of ink (hence size) of that dot.   Not only that, but few inkjets use a traditional halftone screen with a regular halftone dot pattern.  Many printers use proprietary (and undocumented) dithering patterns where there might not be anything corresponding to a traditional halftone dot.

Inkjet printers can use all kinds of tone representations by varying:
droplet size and by that dot size
ink density and by that dot density
and related to the two:
squirting more/less small droplets at roughly the same spot
frequency modulation of identical dots (stochastic)
simulate haftone screens
and use any combination of the methods mentioned.

In offset printing you will see next to halftone screening, also frequency modulated screening, for example with Agfa CrystalRaster.
Offset however can not vary the ink layer density per dot but it can split a tone range in two screens and two print runs, with two different ink densities to give smoother gradations.
Related to litho/offset printing is Collotype (Lichtdruck) and the "dot"  varies in size and density (to a degree) and the screening is non-linear and can have a high frequency.
In conventional rotogravure the linear screen dot (cell) size didn't vary but the ink amount per dot did, later replaced by a mix of halftone and continuous tone screening. Photogravure that led to rotogravure was originally done with an aquatint base that gives it an irregular screening.

So totally new are the tone rendering methods used in inkjet printing not but the technology allowed the use of all methods and hybrids at the same time. One color pixel can be translated to several colored dots that can vary in hue, density, size, shape, distance to one another, the total then is often called a cell. The cell resolution is way lower than the 1440/2880 maximum droplet frequency (dpi) of an inkjet printer. it will be more related to the native (input) resolution of a printer expressed in PPI.


met vriendelijke groeten, Ernst


Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/

[Sven W]

Ernst,
Excellent.
You are the Wiki of LuLa 

/Sven

[Ernst Dinkla]

Sounds better than LuLa of Wiki, for Dutch men at least.


met vriendelijke groeten, Ernst


New: Spectral plots of +250 inkjet papers:

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm

[Sven W]

 

[DRoss]

Even a well known expert photographer, Apple Certified Trainer of Aperture, Photofocus.com blogger Scott Bourne has very little understanding of PPI.

Resolution (PPI) Vs. Dimensions and how they influence file size.

http://bit.ly/ftVCdo

[mfryd]

Even a well known expert photographer, Apple Certified Trainer of Aperture, Photofocus.com blogger Scott Bourne has very little understanding of PPI.

Resolution (PPI) Vs. Dimensions and how they influence file size.

http://bit.ly/ftVCdo

I believe the point he was trying to make was that the setting of the DPI field in the JPEG file is irrelevant.

For most purposes he is correct.  Most workflows ignore this setting, and it has no affect on the printed output.

The DPI field does confuse a lot of people.  It tends to get used when a novice use wants to know an image's size in inches.  The user almost always gets an incorrect answer as the DPI field in the file is probably set to 72, which isn't optimal for print work.

[Ernst Dinkla]

An image file doesn't have a fixed size. It is sometimes easy however to have a "size" tag in the image data. Qimage recognises the PS "original size" in Tiffs and Jpegs saved from Photoshop. It is easier to ask a customer to set the image sizes in PS  than to get a list of images with the sizes next to them. No need to resample, just change one size entry in PS's Image Size, aspect ratio stays the same, PPI just adapts. Save as Tiff or Jpeg. That "size" tag isn't fixed to 72 PPI, the resulting resolution can be as odd as 387,17 PPI.


met vriendelijke groeten, Ernst
Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/