For anyone just starting with digital images, or having trouble getting started, here is an review of the first basics we need, about how to USE our digital images, about how to resize them for viewing them on the video screen or for printing. This is about those first basics of resizing images (i.e., the necessary steps to be able to USE your images). Please appreciate that this can be compared to learning to drive... sure, there are a few details, how to operate the car, and the rules of the road, etc. But you learn it one time, and it's not difficult, anyone can do it. And then you can use the skill for the rest of your life, when knowing how will make a huge difference.
You just gotta know this too. Doing anything at all with digital images REQUIRES knowing at least this much. It is quite easy, if you let it be (if you will simply think "pixels"). If your complaint is that the photo lab does not print all of your picture area (cuts off heads, etc), then you really need to learn this easy material. We all have to know.
Images do have a shape, and the image and print paper sizes are often NOT THE SAME SHAPE. For example, we simply cannot make a 4x6 inch image fit 4x5 inch paper, not without losing an inch off the end, perhaps the top off of someones head. This fact is of course perfectly obvious, but we tend to forget that images also have a shape.
Aspect Ratio is the shape of things - the shape of the image and the shape of the paper. The rectangular shapes may be long and thin in various degree, or might be more nearly square. Aspect Ratio is the simple ratio of the two sides - if the long side is 1.5x the short side, sides are the ratio of 3 to 2. Images from a DSLR camera (and also from 35mm film) are aspect ratio of 3:2. We can enlarge it to any SIZE, but this means for example that the SHAPE of the 3:2 uncropped images will enlarge to print 4 x 6 inches, 5 x 7.5 inches, or 8 x 12 inches. This shape will print "as is" on 4x6 paper, but other paper sizes probably all are different shapes.
Aspect ratio is only critical when matching to one printed paper size, or maybe to full screen monitor shape. Only one ratio fits another shape. And since many shapes exist, no one ratio number is very important, except for your current match, when it is all important.
Size and Shape are different things. Cropping can change the image shape (to fit the paper shape). When we enlarge the image size, it stays the same shape. A 4:5 image can fit 4x5 or 8x10 paper, but not 4x6 or 5x7 paper.
Photo images from a compact or phone camera are 4:3 shape (long side is 1.33x the short side, 4 to 3 ratio). It would fit a 4:3 paper shape, except 4:3 is NOT an available paper shape. So these images suffer a similar problem of needing to be cropped to fit the desired paper.
Photo images from a HD camcorder are 16:9 shape (long dimension is 1.77x longer than short side, 16 to 9 ratio). However, 8x10 inch paper remains 4:5 shape (more below). Not a big deal, we can simply fix it (crop the image to fit the paper shape).
Too much detail, but specifically, here are some common cases of what happens when printing without cropping (which you may have seen). Chart shows the enlarged image size, in contrast to the paper size.
|Uncropped but Enlarged|
Photo Printing Situation
|4x6 paper||5x7 paper|
|8x10 paper||Color of Result Box|
|Images from 3:2 DSLR|
Long side scaled to fit paper
|4 x 6||4.67 x 7||6.67 x 10|
End is cut off
Side is cut off
Side is not filled
End is not filled
|Images from 3:2 DSLR|
Short side scaled to fit paper
|4 x 6||5 x 7.5||8 x 12|
|Images from 4:3 compact or phone|
Long side scaled to fit paper
|4.5 x 6||5.25 x 7||7.5 x 10|
|Images from 4:3 compact or phone|
Short side scaled to fit paper
|4 x 5.33||5 x 6.67||8 x 10.67|
Most printing labs (where we send images to be printed) set up their machines to fill all of the paper. They won't underfill it, they will instead cut something off if necessary. If you order 5x7, you will get 5x7. This means if you print your images without first preparing them (cropping them to the correct shape to fit the paper ordered), you will see surprises about parts of your image cut off, not showing in the final print. This is nothing new with digital cameras, film was always the same thing, simply not the same shape as the print paper. However back then, there was a human operator watching and making decisions and adjusting things for film images. Digital machines are more automatic (inexpensive prints), but digital does make it easy for us to crop it right first. Some online photo printing web sites offer a crop tool there, and will warn when our image is not the same size as the selected paper. And some processors offer an option to "Print Full Image", meaning, they won't crop any off, but will instead leave blank white space (borders) where it doesn't fit the paper.
But the paper size itself is going to crop it, a different shape simply will not fit. But of course the point is that when we do it ourself, then we see it, and we can judge and decide ourselves which portion of the image is to be cropped away... probably we may choose the edge away from the head we want to save. :)
When doing anything with digital images, the first question is "What size is the image?" Digital image size is dimensioned in Pixels. Pixels is what it is all about, and digital is very different than film. If any mystery about pixels, here is a short primer: What is a Digital Image Anyway?, or a more detailed summary at Pixels, Printers, Video - What's With That? Once we accept that pixels actually exist, it's all quite easy.
Resize is a term too vague and ambiguous, it has no specific meaning until we say what it means. There are three very different ways to "resize" an image, and all three have very different meanings and results.
Cropping discards those trimmed pixels, making the image pixel dimensions smaller, but primarily, it changes the scene included, and often the shape too. Different paper sizes (4x6, 5x7, 8x10 inches, are each a different shape - more below in Crop section - therefore we also often crop to make our image shape match the paper shape. Our camera always makes its images of the same one shape (aspect ratio, which is width:height), but our intended use often needs other shape(s), to fit it to the printed paper size or viewing screen size. And frankly, a little cropping often improves the composition of many images, removing distracting or uninteresting blank nothingness around the edges, concentrating the actual subject larger (zooming tighter, so to speak). More detail below, at Cropping.
OK, an example before we get into how to do it. This was a D800 camera image, 36 megapixels, 7360 x 4912 pixel dimensions. To show it here on the web screen (our screens are no more than 2 megapixels size, and many are not even that), it was resampled to an arbitrary 500x333 pixel size, 0.167 megapixels. And by the way (a different subject), do note that even this small image is still quite enlarged here, because the lens image on the camera sensor was much smaller. Now perhaps about 5 inches wide on some screens here (screens vary), but the FX sensor image was only 1.4 inches wide, a DX sensor image is almost 1 inch wide, and a compact image sensor might be less than 1/4 inch wide. Like film sizes (which are generally small too), that's a considerable enlargement. But here, this example is about other properties, like shape.
Trying to emphasize the difference between size and shape.
2:3 is longer and thinner shape, 4:5 is shorter and wider.
If we wanted to print an image that is 2400 pixels tall, then:
Scaling is sometimes called Resize, and Resample is even sometimes called scaling, (not really unreasonable), so the terms can be questionable (what they actually mean in the given usage). My definition of Scaling is about changing the size the image will print on paper (inches), specifically WITHOUT any pixel resampling. It is about declaring the dpi number, in preparation for printing a certain size on paper. This is by far the simplest operation, but sometimes a bit harder to grasp it.
Printing - Before getting into resizing details, first some reasons for them. Printing at home is different than sending the work out.
Saying again, if you send the printing out and order 5x7 inch prints, it does not matter what the actual dpi number in the image file says, because they will rescale whatever you send to print the 5x7 inch print that you ordered. And the paper will crop it, one way or another. Your image dimensions (the resample) should be sufficient so that they necessarily will compute a dpi number around 250 to 300 dpi... around 1500x2100 pixels for a 5x7 print at 300 dpi. Sending more pixels cannot help. Many fewer can be a problem - low resolution, but ± 10% or 15% difference on dimensions and resolution is not an issue. We like to send 300 dpi, but in fact, almost all of these Fuji or Noritsu type machines are set up to print at 250 dpi. Good results either way.
Photoshop for example, if you saved the image with the File - Save As menu, it does embed the dpi number that was set in the Image Size dialog, and will default to print that size. But if you used File - Save For Web, it strips out the Exif, so there is no dpi number saved, and Photoshop will then assume 72 dpi, which won't be helpful. There probably is still an option to Scale to fit media size (the selected paper size) - you just need to insure that resulting resolution will up near 250 or 300 dpi then. But if that file number has been scaled to say 300 dpi, it will scale the image to cover one inch for every 300 pixels, which is what determines print size. A 3000 pixel dimension at 300 pixels per inch will then print 10 inches (this is called scaling).
With either method, a little of our attention first will be naturally be necessary to insure the desired results (sections below). If we want 8x10, we need to crop it to 8x10 shape, and resample it and scale it to 8x10 size. This is not probably going to happen unless you do it.
A printing calculator might be of some help, at least to see the scaling concepts the first time. But the photo editor Image Size tool will do the same calculations, and will do all you need (below).
My own notions about how to prepare our megapixel size camera images for printing are this:
Steps 1 and 2 can routinely improve our image tremendously.
When we print in photo editors (for example in Adobe Elements or Photoshop), we specify paper size, and the default print size in inches is computed from the pixel dimensions and the value of the dpi number stored in the file (3000 pixels at 300 dpi, then 10 inches, which hopefully we prepared for this printing). But we can also print it a different size. This is also scaling. Scaling is changing the dpi number so it prints a different size. Or if you send the pixels out to be printed, and specify 5x7 inches, they will scale it to 5x7 for you. But the printer device itself needs that dpi number when it prints images (which is the only use for this dpi number).
Any cropping or resampling changes are definitely seen on the video screen too. However, scaling does NOT affect the image seen on the video screen in any way. We will see no change at all, because video totally ignores any dpi number and shows pixels directly. But scaling will print a different size on paper if the dpi number changes. More detail below, at Scaling.
OK, here we go.
Adobe: Above is the Photoshop Marquee tool (the "marching ants", like marching movie theater marquee lights - you have to see it). The Marquee tool and there is also a "Crop tool" too work the same in Adobe Elements and Photoshop (both tools in both programs can provide the Fixed Ratio below). This is about both programs, and about most programs. And there are other ways too... I often do this crop in Adobe Camera Raw (select its Crop tool, then right click in the image to see the aspect ratio menu).
There are usually two purposes for cropping.
The Marquee tool is the rectangular icon symbol selected and marked here in yellow near the top of the Photoshop tool bar (right click it for other shape options). Then when this tool is selected, we can simply draw a rectangle on the image, by dragging the mouse over the image (dragging: with left mouse button held down). Then menu Image - Crop will do the crop, leaving only the area in this marked box.
If we also specify a Fixed Ratio in the tools top menu (like 4:5 here, which is marked yellow - Note this is A SHAPE, NOT A SIZE), then any crop box we can draw (of any size) can only be this 4:5 shape, which will fit 4x5 or 8x10 paper. You can see the two different shapes here, right? The overall image is aspect ratio 2:3 (taller narrow shape), and the marked rectangle is 4:5 (shorter wider shape). The point is, you can see and adjust what you are cropping, so that the parts you want will exactly fit the paper shape.
The really BIG trick in the Marquee tool is that as you draw the rectangle, and get it the size that you want, then keep holding the mouse button down, don't let up on it, and then you can hold the keyboard space bar while continuing to hold the mouse button, and then moving the mouse will simply move the rectangle around on the image, to exactly where you want it. Otherwise, you have to start over again. Works very well, but this is much simpler in Adobe Camera Raw crop, where you can just easily do whatever you want to do.
When crop area is selected (drawn to include the image we want to include), then the menu Image - Crop will trim away all else, leaving only the area we marked to be included. Save it to a file, and we have it. Use "Save As" to specify a new file name - Do NOT "Save" to overwrite your only copy of the original, if you expect to ever need it again. The operation is not recoverable (except on Raw images).
Another way - The crop tool in Adobe Camera Raw (including Lightroom, Photoshop, and Elements) is even easier. Just simply mark a rectangle on the image with the mouse. The rectangle shape is forced to match the selected aspect ratio (you can right click inside that rectangle to select other aspect ratios). Then you can grab the corners of the crop frame to drag it to be other size, and you can just scoot it around to frame the area you want. Couldn't be easier.
When we select a tool on that vertical tool bar, we get a secondary menu bar, under the main menu bar. This secondary menu bar has options which apply to the tool we have selected. Here, the Marquee tool bar shows Style, which selects Normal by default. Normal means we can draw any shape box we wish. My choice above was Style - Fixed Ratio, and I specified Width 4 and Height 5, which means any box I draw will be forced to be this shape (not size, but shape), same shape as a 4x5 or 8x10 inch paper. In this tool, if we wanted 5x4 (landscape instead of portrait orientation), we have to specify 5x4, but the left/right arrows between Width and Height will swap the fields. NOTE there are no units in Fixed Ratio. This Aspect Ratio is the simple ratio of the dimensions, and 4x5 is about the shape, not about the size. The idea is to match the "shape" of the paper we will print on. The size will be the pixels remaining after drawing the box, whatever they are (we will fix that in next step below). A minimal amount of practice will make this be clear.
After the marking the crop, you can move the selected area by selecting the Move tool (icon above the Marquee tool), and then dragging it. To change its size, the Marquee tool requires redrawing it, but other tools allow us to drag one edge. You can remove the marked rectangle several ways, usually just clicking someplace else on the image removes it, or the menu Select - Deselect. There are of course other options, but these are the basics.
Yet another way - Photoshop has the Camera Raw crop tool, and has the Marquee tool above, and also has another actually called Crop tool. In it, you can move the cropped frame around on the image, easier to use. Photoshop CS6 has removed the Resolution field from it, which removes my objection and warning. I understand there is some controversy about it, and I don't know any final decision. But CS5 and earlier had it. So a warning - about the formal Photoshop Crop tool before CS6 . It's useful, it has additional powerful options, and does all that too, but it has one feature to watch out about. It's menu bar also shows image width, height and printing resolution. If Width and Height are entered, it is aspect ratio, and becomes the only shape you can mark. That's great. But if resolution is also entered, it will resample to honor it, so that width and height become actual inches on printed paper. Users enter a number because the field is there, not desiring any resample, but it DOES resample. If using this actual Crop tool, and if you enter a Resolution value there (like 300 dpi), then it will crop as marked, and then the Crop tool WILL also RESAMPLE your image, to provide the number of pixels to print the size and resolution that you specified. If that's what you wanted, great. This can combine this Cropping step and the next Resampling step too, meaning, if you want 8x10 inches at 300 dpi (2400x3000 pixels), just say so, and that's exactly what you will get. It is good to be aware however. If you merely want to crop, then simply DO NOT enter a resolution value in the Crop tool. Just leave Resolution blank if only cropping, so it will not resample. Or, I instead just use the Marquee tool above, which simply marks a selection (to be cropped), and will never resample. The Image Size dialog resamples.
Said again: Crop tool will resample if you enter a dpi value (might resample to be smaller, or to be larger, which is good if you intended that, but you should not be unaware). If you use the Crop tool, but merely want to Crop, then you must leave the dpi field blank, and then the Crop tool will only crop to shape, and will not resample. Playing with it a little will quickly show this.
Aspect Ratio is the "shape" of the image - the simple ratio of the images long side to its short side (in pixels or mm or inches, the ratio is the same). For example, 4x6 is long and thin or 8x10 is short and wide (Aspect Ratio). It is naturally important when printing images on paper that the image shape (aspect ratio) matches the specified paper shape (its aspect ratio). Otherwise the image shape will not fit the paper shape exactly. Then most likely, the photo lab will print it so that it fills the paper, but some of the image is cropped away, missing off the edges of the paper. It is of course far better if you see and plan and do this crop yourself, instead of the print lab doing it randomly (We have photo editor tools to make this be easy work). Another possibility if printing at home is that some of the paper area is left as blank white space (if shapes don't match).
This is just saying, a 4x6 image cannot fit 4x5 paper, simply Not the same shape. This is not about Size, size can be enlarged as necessary, but Shape can only be cropped. It is a very simple concept (like square pegs in round holes), but not all beginners seem to take notice yet. A 2:3 image will enlarge to 4x6 or 8x12, but it will not fit 8x10 (unless first cropped to 4:5, one way or another). It is a complication that most paper sizes are all different shapes (except ISO metric paper, but which is still different than our images).
Aspect Ratio simply compares the image width and height dimensions, for example 4x6 inch paper is ratio 4:6, which lowest common denominator reduces to 2:3. Or a 3000x2000 pixel image is 3000:2000, which is reduced to 3:2 (same ratio as 2:3, just rotated, but still fits the paper). Or the aspect ratio can be described as the division into the larger dimension, so 2:3 is 3/2 = 1.5, and the ratio compared to "1" as 1.5 to 1... same shape as 2:3.
8x10 inch paper is 8:10, which is reduced to call it 4:5. The big point is that 2:3 and 4:5 are simply different shapes of paper. Obviously, we cannot print on a different shape paper without losing some area (cropped away, off the paper).
Said again, 4:5 aspect ratio can be scaled to fit 8x10 inch paper, but a 2:3 image will require 8x12 inch paper (else on 8x10, something is cropped off and lost). Or, the idea is that you can do that cropping yourself first, intelligently, your choice of what is cropped.
When printing the image, you often need to crop it first, to fit the paper shape. Because, 4x6, 5x7, 8x10 inch paper are all different "shapes". 4x6 is a more elongated shape than 4x5 (so the same one image cannot possibly fit both shapes of paper). 4x5 or 8x10 shapes are a little wider, so to speak. Shape and size are different properties. Images can be scaled or resampled to different sizes, but the shape follows in proportion. 8x10 can print on 4x5 paper, but not on 4x6 paper (because 8x10 shape is 4x5). If the image and paper are different shapes, the paper will necessarily crop it - some of the image will be off the paper and lost, or there will be blank unfilled white space - so the correct shape is important. It seems reasonable and necessary to decide this yourself about how to crop your image, before you print it. And frankly, a little tighter cropping often helps many images, like this one above, making the remaining subject area larger and more important.
This Cropping step above shows how to crop the image to fit the paper.
My notions: Aspect ratio is technically width:height, and width:height is reversed by rotation. The paper "shape" is what it is either way, but there is also Landscape and Portrait orientation of the paper, which swaps width and height. Landscape orientation is the long dimension turned sideways, like say 8x10 paper turned to 10x8, matching the normal camera orientation to take a scenic landscape picture. Portrait orientation is the camera turned up on end, like say 8x10, like most portraits are done. We may rotate the paper to use it, but we rarely hear the terms said as 10x8 or 2:3 or 11x8.5, etc. It's more what we are used to saying. 8x10 paper is 8x10 paper. Anyway, popular aspect ratios like 4x5 or 5x7 are usually said as portrait orientation, but 3:2 seems always be landscape. The numerical divisions (1.25 or 1.4 or 1.5 to 1) are usually landscape, same as the cameras normal mode... because the long side divided by short side is a number greater than 1. So note that the numbers are different, 5x7 and 7x5 do divide as different ratios, 1.4 vs 0.714. We get used to this. :)
Some common Aspect Ratios are shown. (ratio of the two dimensions, which is "shape"). When the ratio is say 3:2, the sides compare as 3 to 2. Same ratio giving W/H:1 of say 1.5:1 (3/2 = 1.5), means the images long side is 1.5 longer than the short side. This is a "shape", not a "size".
The formats are ordered here from a square shape to a long and thin shape.
|W : H||W/H to 1|
|1:1||1 to 1||Square|
|4:5||1.25 to 1||4x5, 8x10, 16x20, 20x25 cm (inches or cm, doesn't matter).|
|11:14||1.273 to 1||11x14, 28x36 cm. Was an old Kodak sheet film size|
|8.5:11||1.294 to 1||American Letter size, American A size|
|4:3||1.333 to 1||Most compacts, phones, and non-widescreen monitors|
|5:7||1.4 to 1||5x7, 13x18 cm. Also 2.5x3.5 inches wallet size, which is 1/4 of 5x7 size|
|Metric ISO||1.414 to 1||Metric paper sizes, A2, A4, A6, etc. √2 :1|
|3.5:5||1.429 to 1||9x13 cm. Many old snapshot photos, also 9x13 cm, or half of 5x7|
|2:3||1.5 to 1||4x6, 8x12, 10x15cm, 35mm film, and most DSLR cameras|
|11:17||1.545 to 1||American B size, also Ledger (landscape) or Tabloid (portrait) size|
|16:10||1.6 to 1||Some wide LCD monitors, near Golden Ratio 1.618|
|8.5:14||1.647 to 1||Called Legal size, law profession|
|16:9||1.778 to 1||HDTV 1920x1080, 1280x720, and camcorders|
Note: Wallet size can vary slightly. Traditionally 2.5x3.5 inches (measure your wallet picture insert sleeves), which is 1/2 of 5x7 dimensions - meaning, 5x7 images will scale to fit wallet size with no cropping - four can be cut from 5x7 paper, but borders may make them slightly smaller. Some are 2x3 inches. My experience is Walmart advertises their wallet size as 2x3 inches, but they measure close to 2.5x3.5.
Metric paper sizes A2, A3, A4, etc, are each cut or folded to be half of the previous, but all are same aspect ratio 1.414. Square root of 2 is a magic property this way.
American paper sizes, A 8.5x11, B 11x17, C 17x22, D 22x34 are all double size of the previous, and they also cut or fold in half, but with two aspect ratios alternating every other size.
The point is, these shape differences are the reason people complain that the photo lab didn't print all of their picture area. To prevent the lab from cropping your image in an surprising way (but it is expected, the image shape must fit the paper shape), you should crop to 2:3 shape BEFORE you order 4x6 inch prints, or to 4:5 shape BEFORE you order 8x10 inch prints. Then what is cropped is your own choice. If you don't, and if you send the lab images that do not fit the requested paper size, the lab will have to enlarge it enough to fill the paper, cutting off a little of your expected image. Or sometimes you can request they just print it leaving blank white space where it does not fill. It is best if you tend to this before you send it, cropping the way you want it so that the image shape fits the paper shape without surprises. Note the obvious, in these "red" images above at right, we can of course move our own cropping selection to be at or near either edge, to better center our actual subject in the final area.
Most paper sizes are each a different shape (except metric is consistent, all actual ISO metric sizes (A, B, C series... A4, A6, etc) are the same shape) but few paper sizes are the same shape as our images. In the old days, a human operator might see and select a better print cropping, but the machines are fully automatic today, so forget that, at least for discount price. This is something you surely want to see and choose first, your way, instead of letting the printer do it automatically by machine, their way (often unseen by human eye).
The image above has been cropped now, to show the image that I decided I wanted to show, and also to fit paper shape of 4:5 aspect ratio. This step gets it to 8x10 inch size when printed.
Also shown is the Adobe Image Size dialog box (same in Photoshop or Elements ). It is menu Images - Image Size, or keys CTRL L pops it too (speaking Windows). Most serious image editor programs necessarily do about the same thing, and work about the same way. And BTW, the Best Help available for Adobe stuff is simply to do a Google search for the words you want to know about.
This tool does two very different resize functions: Resample, or Scale for printing. The mode switch on this tool is the Resample Image check box (bottom), checked so it will Resample. If NOT checked, this box will instead Scale for printing (next part below), and then, the top portion of the box (pixel dimensions) is grayed out, image size cannot be changed. But if checked, it will resample, and the pixel dimensions are changed.
Resampling changes the image size, specifically, it changes the dimensions in pixels to be other dimensions you specify, other new pixels. Several reasons why and when we do this, but our camera images have become larger than most uses need, so we generally need to make them smaller, to be the image size needed for our purpose. For example,
There are two ways we can use this box to resample:
(8 inches x 250 dpi) x (10 inches x 250 dpi) = 2000 x 2500 pixels.
The resample tool is a calculator that will do this multiplication. Just enter 250 dpi and 8x10 inches, and the pixel dimension numbers change so that this new image size is what you get - designed for this specific printing purpose.
Be careful that your numbers do not increase the original size in pixels (which is interpolation, and not a good thing... often better then to uncheck Resample, and just scale the "pixels per inch" number to fit at a lower resolution (see Scaling below).
After the first crop above (done for shape and content), this tool shows this image size now is 2514x3143 pixels (4:5, or 3143 x 4/5 = 2514). Resampling smaller is a drastic operation, which works great, but there is no going back after the excess data is discarded. First archive your largest original data somewhere unmodified (Raw is great for this).
Note the number on the top line of this tool - the 22.6M is the data size in megabytes, 22.6MB. This is not file size, and it is not image size. This is size of the image data when uncompressed and open in computer memory. It is a computed number, from the image size (pixels). For 8-bit images (24 bit color), this number is always 3 bytes per pixel (it is the RGB data for each pixel). Or 6 bytes per pixel for 16-bit images. So any generic 12 megapixel image is 12x3 = 36 million bytes of RGB data. True of any/every 24-bit RGB 12 megapixel image. But this one is now cropped to be smaller, 7.9 megapixels now (2514x3143).
This 22.6 MB number is close, but not exactly (2514x3143 x 3) bytes. Megabytes is an odd situation. One megabyte (1024x1024 bytes) holds nearly 5% more bytes than one million bytes (1000x1000), so there are about 5% fewer megabytes. See More about that. But this size in memory simply depends on the pixel dimensions, as shown, and it is the size of our data, usually three bytes per pixel.
Image data size is not file size. JPG compression makes the data smaller while in the JPG file. File size is variable with the JPG Quality option (High JPG Quality is larger files and better images, Low JPG Quality is smaller files at the cost of worse images). File compression size varies with image content too - featureless image areas (walls, sky, etc) compress smaller than detailed image areas (tree leaves, etc). We speak of JPG Quality losses, which means the addition of JPG artifacts, see More about JPG. The JPG file size might be tolerable quality if around 1/8 to 1/12 of image data size (very vague numbers, but larger files are better images). The image is the full original size again when opened and uncompressed into computer memory (same pixel dimensions), but quality might suffer from JPG compression. The top line of the resize box shows this uncompressed data size, in MB. This is computed at three bytes per pixel (for 24-bit images). That is simply how large our image data is (JPG files tend to give us other false notions about it).
In this Image Size box - with the Resample Image check box checked - we can Resample to change image size (pixels), done in a couple of ways:
Currently above, this tool's Document Size center portion is showing that this image will print 8.38 x 10.477 inches at 300 dpi. This is simply because:
2514 pixels / 300 pixels per inch = 8.38 inches
3143 pixels / 300 pixels per inch = 10.477 inches.
Enter any other numbers, either inches or dpi, and you will see different results for that goal.
It is just this simple - pixels per inch. For example, if you intend to print 4x6 inches (2:3) at 250 dpi for Walmart (the image will first have to be 4x6 aspect ratio (2:3) for it to come out this shape), you can enter 4x6 inches and 250 dpi in this center portion (Document Size). Any number you type in this Document Size section WILL RESAMPLE THE IMAGE to change the pixel dimensions in the top Pixel Dimensions section. 4x6 at 250 pixels per inch will become a smaller image, it will come out 1000x1500 pixels, and 4.3 MB, but it will be right to print 4x6 at 250 dpi (because 1000 pixels / 4 inches = 250 pixels per inch). You can change either resolution, or dimensions, or both, and it will resample to give you the new pixels for the image size that you specify. Then press OK, and you've got it. If you Save to overwrite the original file, then the original data is gone, unless you archived a copy first, so think that over first - save it to a new file name. Just experiment a few minutes entering different numbers here, and you will get the idea. This is just a simple calculator, but it DOES change the image size (pixels) to be what you specify.
Dpi and inches are related, tied together for printing, according to the number of pixels available. The inches exist only on the print paper (not in video). Dpi is just an isolated number stored in the image file, just a number, used as instruction to the printer, to space the pixels on paper. Dpi is not a property of the image data. Your 12 megapixel JPGs may come out of the camera with this number marked maybe 240 dpi (arbitrary, varies with camera brand), which ONLY means that it then indicates a print size about 12x18 inches. But of course, it is just a made up number - the camera has absolutely no clue or concern how you might print it later. It is just a dummy number - its only purpose is to "not be blank" (so the computation for indicated inches does not blow up). It has no meaning at all, not until you make it show the way you actually want to print it. It is NOT about creation, and it is not about image resolution. The image size and detail depend only on the pixel dimensions, and not at all on this fake dpi number. Yes, it is called resolution, but it is only about printing resolution, how the printer will print it, when you print it, and is NOT at all about image resolution. You can change this resolution number to any number you want (but this Resample box will resample then, where Scaling below will not). DPI is not a property of the image data. Image size is dimensioned in pixels (I may have mentioned that?) Image detail depends on how many pixels we have, and is not affected by this number (until we print it, then dpi only affects how the printer will space those pixels). Video systems do not use this dpi number in any way. Video screens simply show pixels directly. etc, etc, the rest of the site here is about that stuff.
You do want to pay attention to be sure you are resampling smaller, and not resampling larger. For printing, instead of resampling 20% larger, which just makes the image softer (nothing can add detail from the original scene once out of the camera), then it is surely better to simply print it scaled to 20% less resolution (pixels per inch), which increases printed size the same, but leaves the pixels unaffected (which is Scaling, next below).
Bring up this tool and experiment by typing some numbers. It should become very clear, very fast. It is just a simple calculator, about pixels per inch.
The last list box at bottom is the resampling algorithm to use. For photos, Bicubic Sharper is a fine choice for reducing the image size. Bicubic Smoother can be better for increasing image size, but increasing size is generally not a great option. Note that all images need a moderate bit of sharpening after resampling (all new pixels then). Any sharpening done before resampling is generally pointless, as those pixels will be replaced with different pixels... Sharpening ought to be done last, the final operation. Large images for printing need a bit more sharpening amount than smaller images for video. Very small images cannot stand much sharpening. We sharpen for the viewing goal.
This option is ONLY to prepare for printing the image, to cause it to print the size we want on paper. Whereas... the Resample option above is about changing the pixel dimension of the image, for any purpose, but for example, maybe to make it display smaller on a video monitor. Scaling is very simple, all we do is change the dpi number, specifically without changing any image pixel in any way.
Video and printing are very different. Video monitors simply show pixels directly. If we want to see a different size (maybe the image is too large for the screen), then the image must be resampled, to create a different or smaller image size, the pixels of which are always shown directly on video. Whereas printers operate very differently - printers fill inches of paper by spacing the pixels at different "pixels per inch" rates. Scaling simply changes that dpi number, so the same pixels will cover a different size in inches (on paper). The scaled image will appear unaffected on the video screen, which ignores the dpi number, and shows pixels directly.
Scaling does NOT change image dimensions (in pixels). It never affects any pixel at all, and it has zero effect on the video screen. It is only for preparing to print, and technically, scaling only changes the Resolution field "Number" (pixels per inch), which is just a number. Some new dpi number just causes the printer to compute to print the same pixels to cover a different size in inches on paper (pixels per inch). The dpi number simply "scales" the printed size on paper, by changing the printed pixel spacing to make the SAME pixels cover a different dimension of paper, in inches. The pixels stay the same, totally unchanged. 3000 pixels printed 300 pixels per inch will cover 10 inches of paper, or printed 200 pixels per inch will cover 15 inches (same 3000 pixels). Scaling is very reversible, can be done at will, since there was no actual image change at all (just this one dpi number stored away separately - it is just a number).
We initially cropped this image to 8x10 inch shape (4x5 aspect ratio), so now lets suppose now we want to print it 8x10 inches. If we just feed this image (in first part above, after cropping) to our printer, what this said is 8.38 x 10.477 inches at 300 dpi, and that is how the printer will print it (It is how our local home printer will print it. If we take to Walmart or such, they will print it any size we ask for, i.e., they will scale it for us - they don't look at the dpi value).
Here, same tool, but I unchecked the Resample Image check box, so now it cannot/will not resample. This selects Scaling mode. The top fields become grayed out, the pixel dimensions cannot be changed. All it does is change the separate dpi value, but there is this specific option for it, this is how we do it (it saves the new number in the file, if we save the file). Notice again, the top part (Pixel Dimensions) is grayed out now, we CANNOT change the pixel dimensions in any way. We are not resampling, and we cannot change the image size. All we can do is to just change the dpi number, which changes the printed size, after it is on paper. The purpose is to scale the image size (pixels) to the paper size (inches).
However, we can change the dpi value, or change the printed dimensions in inches, and then the other one will track it. It is just a calculator. I also typed 8 inches for image width in the center at Document Size. This is printing size. The 10 inches automatically changed (because this is the aspect ratio), and the 314 pixels per inch is just the simple calculation: 2514 pixels / 8 inches = 314 pixels per inch. It is still the original pixels, but this dpi value will cause our home printer to print it 8x10 inches.
This 314 dpi is close enough, it is not a critical or exact number. For best printing quality, we would always like it to be around 250 to 300 dpi for printing. Within 10% or 20% is fine (when necessary, when that is the pixels we have), and will be undistinguishable. The places you send it out to for printing normally have capability around 250 dpi, and you would like to have that much. If you don't have enough pixels, then a 150 dpi result is possibly acceptable (not terrible), especially for large prints which are not viewed too closely. My own goal is to provide 300 dpi, "just to be sure"... but 400 dpi is pointless, won't help.
So if we wanted to print this one 4x5 inches, this 2514x3143 pixel size is simply far too many pixels this time. Scaling to 4x5 inches will come out 2514 pixels / 4 inches = 628 pixels per inch, which is simply ridiculous for printing color photos (but is fine and good if for line art). Realistically, printing at 250 to 300 dpi size is all that can benefit the printer (any printer, for color photos), so resampling smaller is often necessary and prudent. It would print if we left it scaled to 628 dpi, but the file and upload would be 4x larger than necessary, for no reason at all.
And when we print, generally the printer driver will use the dpi number to size or scale the printed image. However, many printer drivers today will also allow us to specify the printed size there too. This is the same scaling operation (changes value of dpi to fit the image to the page size). We still need to have the shape right however.
Note this dpi resolution number is just a stand-alone number, which does not affect the image pixels at all (unless we also resample, but this is speaking of scaling - the Resample Check Box is Unchecked). And it only has meaning at the instant of printing on paper. It is arbitrarily chosen to decide print size scaling, and it is unused at all other times. Images viewed on the video monitor do not use this dpi number in any way, therefore images on the internet web also simply do not use this number in any manner. The Adobe menu File - Save for Web strips out this number to save a few bytes, since it is not needed on the web (Adobe reports 72 dpi then, because it is missing, not because it says 72). Video systems simply show the image pixels directly, and in the video system, it simply could not matter less what this resolution number is. It is totally ignored in the video system. It is used for printing on paper, to scale the printed image size to be so many pixels per inch.
Bottom line, if the image size is halfway close to proper printing size, there is no reason to modify and affect the pixels. Just leave them alone, use what you've got, and scale the image to print the exact desired size. If the printing resolution comes out near the ballpark of 250 to 300 dpi, you're good. This Document Size tool is just a simple calculator, which computes: pixels / inches = pixels per inch. And when you click OK, it stores this dpi number in the file, to retain it for printing. Your local printer device uses that number to size the image pixels on paper, so this is how you make it print a specific size in inches. The dpi number is stored in the file (is just a number), but the inch numbers are not stored. There is zero effect on the image pixels if we simply change the dpi number. Then the unmodified existing pixels just print a different scaled size image on paper.