How big is a pixel?
Everything You Always Wanted to Know About Pixels – But Were Afraid to Ask
Pixels and derived terms such as ppi/dpi and resolution are probably among the most misunderstood things in digital photography. The major source of confusion is the fact that some ppi-number is embedded in the image EXIF metadata by most digital cameras. This gives some people the idea that pixels do have size and that print resolution is somehow a basic attribute of the image file. This is, of course, nonsense. Size in inches is not really an attribute of a digital image until it is rendered on some output device, so there is no reason one should attach any meaning to the print resolution embedded in the file. But people do, and get confused.
Below in an attempt to give plain answers to some of the most frequently asked questions about pixels and resolution.
Table of Contents
- What is resolution, and how is it expressed?
- My expensive camera outputs images that are only 72 ppi! How can I increase its resolution?
- Is the embedded ppi-number useless, then?
- What is my camera's pixel count?
- How big is a pixel?
- Wait! The pixel pitch on my camera's sensor is 9 micrometers. Isn't that a size?
- Wait! My astronomer friend talks about the angular size of her pixels.
- If the the sensor's pixel pitch is large, will that not result in lower resolution?
- So the physical size of the sensor cell or the sensor chip does not matter at all?
- Is a camera with more megapixels always better?
- What is the largest size I can print?
- What is the minimum resolution I can use to produce an acceptable print?
- How do I print at a specific resolution?
- How do I print an image at “100 %” or at its “actual size”?
- I used Photoshop to change the resolution on my image to 72 pixels/inch to match the resolution of my screen. It looks fine on the monitor screen, but when I print the image – it is horrible! What's up?
- I have a nice 400x300 pixel image that I saved from the web. When I make a 4 by 3 inch print from it, it looks bad. What's up?
- The printer/editor at the magazine that is to publish my images said that the images had to be at least 300 ppi. How do I change the resolution to 300 ppi?
- What size images should I use on my website?
- How do I set the ppi-number of an image in Photoshop?
Q 1: What is resolution, and how is it expressed?
Resolution is a measure of the ability to resolve and separate fine details that are positioned close to each other.
Resolution in a photograph, digital or not, depends on a the entire system used to capture and reproduce the image for viewing. Assuming that focusing is accurate and that no motion blur or mechanical problems influence the image quality, the two most important factors influencing resolution in a digital image is the quality of optical system used for capture, and the number of pixels used to make up the image file.
The number of pixels along the height and width of a bitmap image is called the pixel dimensions of an image. In a lot of literature and advertising material, the word “resolution” is used as a synonym for the product of the pixel dimensions – usually measured in megapixels. This implies that resolution is determined by pixel dimensions alone, and leave other important system components, such as the optics, out of the picture. To avoid the confusion this creates, this FAQ will instead use the term pixel count to denote the product of the pixels dimensions. (see Q 4 below for more about this).
The resolution of a digital image when it is rendered on an output device (e.g. paper or screen) is usually expressed as pixels per inch (ppi). Photoshop CS expresses resolution as pixels/inch, and Corel Photopaint, IrfanView, Microsoft Office Picture Manager. Nikon Capture NX2, and Photoshop Elements unfortunately calls this dots per inch (dpi) – but in all cases, it is the same measure.
Note: Manufacturers of inkjet printers and other halftone devices also use “dpi” They do not mean the same thing as in Corel Photopaint, etc. A JPEG photo uses 24 bits continuous tone to represent one pixel. This means that each pixel can have 16777216 different colour shades. To render this, a first approach would be to have a printer with 16777216 different colour inks. Real world ink jet photoprinters have only have between four and eight ink cartridges. To make up the remaining colours they use one variation or other of a technique known as halftoning, dithering or stochastic screening. This is a technique where different coloured dots are printed so close together that the eye blends them together and see an intermediate colour. This means that an ink jet printer needs to use many “dots” to print a single “pixel”.
Q 2: My expensive camera outputs images that are only 72 ppi! How can I increase its resolution?
The ppi-number embedded in the image file by your camera is meaningless as a measure of resolution. The resolution of a digital image is indicated by its pixel count, not its embedded ppi-number. In most cases, when you display your images on your PC's monitor, on the web, or when you print them on a photoprinter, the embedded ppi-number is not even used.
This is demonstrated by the two images below. The image on the left is 72 ppi. The image on the right is 300 ppi. But as long as their pixel dimensions are the same (300x150 pixel), the difference in ppi-number does not matter: Image quality and file size are identical.
Q 3: Is the embedded ppi-number useless, then?
The embedded ppi-number is useful when integrated in the prepress workflow when typesetting magazines and newspapers (e.g. for screen proofing and for page layout). It is indeed used by some high-end typesetting packages intended for such a workflow.
Also, setting this number to an appropriate value may also be useful if you want to preview the image's print size to optimize sharpening and cropping for a specific printer. For more about this, including how to change this number without degrading the image, see Q 13, Q 17 and Q 19 below.
Q 4: What is my camera's pixel count?
A digital camera's maximum pixel count is measured in the number of megapixels (millions of pixels, often written as Mpx) it can record. For example: The grid that make up an image recorded by a Canon Powershot G5 is 2592 pixels wide and 1944 pixels heigh. If we multiply 2592 by 1944, we get 5038848 (slightly more than 5 million). This is why the Canon G5 is referred to as a 5 megapixel (Mpx) camera.
Most digital cameras can be used to capture images at more than one size. For most cameras (the so-called high-speed mode of the Nikon D2x is an exception), the smaller image is created by using software to downsample an image recorded at maximum image size to a smaller size. The field of view (FOV) of the downsampled images is identical to the original, but there are fewer pixels in the smaller images. (See the answer to Q 19 below to learn more about downsampling.)
Unless you need to save space on your storage card for some reason (e.g. you are on a three week safari and forgot to bring spare cards), there is no point using a smaller pixel count than the maximum your camera can deliver. You can always use those extra pixels for cropping, or for making larger prints.
Most consumer digicams have something called “digital zoom”. This is a gimmick where the center of the image is cropped inside the camera, and then upsampled by means of interpolation (for more about interpolation, see Q 11 below) so that the cropped result is the same size in pixels as the original image. While the number of pixels that result from this process indicate that the image has a lot of detail, this is not true. The interpolated pixels created by the upsampling process carry no information, so “digital zoom” will not give any more detail. There is nothing to be gained from “digital zoom” – it is useless. Just forget that it is there. (If you want to “zoom in” on a particular portion of the image, you can always achieve the same effect as “digital zoom” by cropping.)
Q 5: How big is a pixel?
A pixel has no size.
Q 6: Wait! The pixel pitch on my camera's sensor is 9 micrometers. Isn't that a size?
Yes. But this is not the number most people are looking for when they ask how big a pixel is. They are talking about the pixels in their image files and what their size might be in the context of rendering images. The physical size of a sensor sensel bear no relation to that.
Q 7: Wait! My astronomer friend talks about the angular size of her pixels.
Earth scientists, astronomers and others using images from spacecrafts and telescopes may well be able to relate each pixel to the size of a given ground coverage region or angle.
In their jargon, the term “pixel size” is shorthand for the physical size of the plantary surface area or angle of view into the universe that are projected onto a single pixel. This techical jargon, while useful in planetary science and astrophotography where the “pixel size” is used to convert pixels into physical measurements, has little meaning in the context of general digital photography.
Q 8: If the sensor's pixel pitch is large, will that not result in lower resolution?
No. In digital photography, it is just plain wrong to relate resolution to area measured on the surface of the digital sensor. None of us view our photographs by examining the sensor cells at the surface of the sensor.
The normal way to view a photograph is to make a print on paper, or to project it on a computer screen, and to look at this print or projection. In such a viewing context, it is not the physical sensor size that determines the resolution, but (among other things) the total number of pixels available.
Q 9: So the physical size of the sensor cell or the sensor chip does not matter at all?
Yes, it matters – is just does not matter in terms of resolution.
If all other factors are equal, larger sensor cells produces less noise and a usually also a larger dynamic range.
The physical size of the sensor chip also determines the field of view at a particular focal length. A large sensor gives a larger field of view at a given focal length than a small sensor will. Also, a large physical sensor will give the photographer more control over depth of field.
Q 10: Is a camera with more megapixels always better?
There two advantages to having more megapixels: You have the option to make a larger print (see Q 12 to learn about print sizes); and you have greater leeway if you want to crop a portion of the image.
However, sometimes cameras having less megapixels might have an advantage. If too many megapixels are put on a too small sensor, its performance in low light may suffer.
You may get rid of (some of) the noise you get from small sensors with a high pixel count by downsampling the capture to a smaller pixel count. However, when you do this, you also lose the advantages associated with a high pixel count, so nothing is gained.
Q 11: What is the largest size I can print?
That depends on a number of things, including the number of pixels available, the intended viewing distance, and your standards for quality. You can make a large print from an image with a low pixel count if the image will be on a poster on a distant wall that only will be viewed from a distance.
There are some guidelines for working out the maximum print size for a given pixel count and resolution (see the next segment to learn about print sizes), but there is no single “right” answer.
All sorts of imperfections becomes more visible at larger print sizes, so such factors as the presence of JPEG artefacts, noise or dust, lens quality, blur due to inaccurate focusing or camera movement, and general image quality, also need to be taken into account when determining print size.
On the other hand, a low pixel count can, to some extent, be compensated for by upsampling, using a technique know as interpolation. How succesful this will be depends both on the characteristics of the interpolation software used and the type of picture. Images with large areas of uniform colour usually interpolate better than images with a lot of fine detail.
Interpolation is a huge subject, and there exists a number different ways of doing it. To learn more about this, and for a comparison of different algorithms and programs, see our interpolation page and Americas Wonderlands': Digital Photo Interpolation Review.
Q 12: What is the minimum resolution I can use to produce an acceptable print?
As noted in the previous segment, that depends. However, there are some guidelines that can be used as starting point.
For newsprint, 172 ppi seems to the norm. For magazines and glossy prints, assuming normal viewing distances, a print resolution of at least 240 ppi is recommended for acceptable print quality. A perfectionist will say that you should use 300 ppi or more. These are ballpark figures. You need to experiment to find the figure to base your own standards on.
Also, as discussed in the next segment certain printers have an optimal resolution. By making sure you are printing at this specific resolution, you will get the best possible quality from the specific printer.
You can work out the print resolution from the pixel dimensions by taking the number of pixels along one of the sides of image and divide by the length of the same side of your print. For example, if the long side of your image is 3600 pixels, and you are printing at 12 inch paper, your printing resolution is 3600/12=300 ppi.
If you learn how to use Photoshop's Image Size dialogue (see Q 19 below), it will do the math for you.
Q 13: How do I print at a specific resolution?
When you order prints from a commercial lab, you also specify the print size in inches or centimeters. If necessary, the RIP (Raster Image Processor) in the lab's photoprinter will resample your images to fit, and then print it at the specified size. If you are doing your own printing, you can use the printer driver dialogue to set the desired size of the print and the printer's RIP will work out the correct resolution and (if necessary) resample to fit.
I.e.: As long as you make sure that your image has a suffiscient number of pixels to produce a quality print, the printed image will be fine. You never need to specify a ppi-number when printing – the physical dimensions will do. Refer to Q 12 to see how you can determine the minimum pixel count needed to produce a quality print.
However, if you are doing your own post-processing, including image sharpening, you may want to preview the sharpened image on your screen at print size. In that case, it may be helpful to have the appropriate ppi-number embedded in the file. Please refer to Q 19 below too see how you do that with altering the pixel count..
Further: Output devices are optimized for a certain resolution. For desktop inkjets, this tend to be 600 ppi for printers from Canon and HP, and for Epsons it is 720 ppi. The Fuji Frontier used in many minilabs is usually optimized for 300 ppi, Lightjet 5000 needs exactly 304.8 ppi, while the Durst Lambda series of laserjet printers offer a choice between 200 ppi and 400 ppi. However, you should refer to the manufaturer's specifications or ask the lab to determine the optimum resolution for a specific printer.
If the files submitted for printing have a pixel count that does not match (i.e. too low or too high) the printer's optimum resolution for the requested output size, it will – as noted above – be resampled to match. Some printers unfortunately use sub-standard methods for resampling. This step may result in a slight degradation of image quality.
To avoid this, you should make sure your files have the optimum resolution before they are seen by the printer. You can do this with the help og the Image Size dialogue in Photoshop and similar tools (see Q 19 below), or use special programs developed for the task, such as Digital Domain's QImage.
There also exists dedicated RIPs that bypass the printer's built in software and deal with the printer hardware directly. They use proprietary algorithms for interpolation and halftoning, and may produce even higher quality output.
Q 14: How do I print an image at “100 %” or at its “actual size”?
Outside the fields of macro photography and artwork reproduction – where it is sometimes desireable to make prints where objects on the print have exactly the same physical size as they have in reality – the terms “100 %” or “actual size” isn't meaningful in the context of printing digital images on paper.
Q 15: I used Photoshop to change the resolution on my image to 72 pixels/inch to match the resolution of my screen. It looks fine on the monitor screen, but when I print the image – it is horrible! What's up?
72 ppi is a too low resolution for prints.
You should never use the Image→Image Size … dialogue in Photoshop to change the resolution unless you know what you are doing. Your image is destroyed. I hope you have a backup copy.
Q 16: I have a nice 400x300 pixel image that I saved from the web. When I make a 4 by 3 inch print from it, it looks bad. What's up?
The image has too few pixels to print OK. Printed on paper 4 by 3 inch, this is equal to a resolution of only 100 ppi. For a quality print you need more. Refer to Q 12 above to learn what type of resolution you need for a quality print, and how to compute the print resolution from the number of pixels.
Q 17: The printer/editor at the magazine that is to publish my images said that the images had to be at least 300 ppi. How do I change the resolution to 300 ppi?
Usually, there is no need to change anything. What is meant by this is that your image should have a pixel count that shall make it possible for them to print at a resolution of 300 ppi when they print it at a specific size.
To figure it out, you take the number of inches the print is going to be when printed, and multiply that with the desired print resolution (here 300 ppi). This gives you the desired width and height, expressed in pixels. Now, multiply the width (in pixels) width the height. That gives you the number of megapixels you need.
Example, if the printed copy is to be 12 by 8 inches: 12x300=3600, and 8x300=2400. So your image should be at least 3600 x 2400 pixels – that's about 8.6 Mpx.
Note that some professional typesetting packages, such as Framemaker, makes use of the embedded ppi-metadata when laying out pages. If that is the case, check out the answer to the this question to see how you change this without losing image quality.
Note: Sometimes, you may come across requests for images to be “300 ppi” – without any associated physical dimensions. Unless the physical dimension can be inferred from context (e.g. a glossy magazine full page spread), this is nonsense.
Q 18: What size images should I use on my website?
There is not a single correct answer. For one thing, it depends on the layout of your website. If your site is a blog, your images should be scaled to fit the columns size (e.g. 512 pixels) of your blog. If you publish images on a photo album site, you may instead opt for using use the full width of the computer screen for images.
Most computer screens these days are at least 1024 pixels wide and 600 pixels high. But screens as large as 1920 by 1080 pixels are getting more and more common.
My recommendation is that for the web, you resize images to be 1024 by 600 pixels, or smaller. That will ensure that most people that visits your website will be able to see the whole image without scrolling.
JPEG images are always compressed, and the algorithm used to compress them is lossy. However, with moderate compression, the human visual system does not see that a lot of information is missing. The “quality” you select when you save a JPEG file determines the strength of compression and therefore also the size of the saved file in Kbytes (kilobytes). Big files are slow to load and take up more space on your disk. You therefore need to weight size against quality when saving images for use on a website. As a rule of thumb, I find that a good compromise is maximum 300 Kbyte for an image that is 1024 by 600 pixels, and 80 Kbyte for an image that is 512 by 300 pixels.
Q 19: How do I set the ppi-number of an image in Photoshop?
The dialogue to change the ppi-number in Photoshop is Image→Image Size …. When you choose it, you get a dialogue box like the one shown to the right.
Pay attention to the checkbox labeled Resample Image. If this is checked, Photoshop will resample the image when you change resolution to keep the physical dimensions of the image constant. For example: If the resolution set by the camera is 180 ppi, and you use Photoshop change this to 72 ppi. Photoshop will throw away 84% of the pixels. If this is not what you want, make sure that you leave the Resample Image checkbox unchecked.
If you uncheck Resample Image, this dialogue turn into a resolution calculator. You can now play around with the resolution to see how big the image will print if printed at that specific resolution, or you can play around with the width or height, to see what the print resolution will be for that particular size.
If you save the image after changing the resolution with Resample Image unchecked, all the pixels in the image will be left unchanged. The only thing that will change is the ppi-number embedded in the file. Those who are unhappy with the default used by the camera's manufacturer (see Q 2 above) or need to change it because their page layout program uses this number (see Q 17 above), or want to preview the image's print size (select Zoom Tool and click Print Size), can use this function in Photoshop to get a number they like.
As long as you leave the Resample Image checkbox unchecked, you can play around with the ppi-number without accidentally damaging your image.
If you check Resample Image, Photoshop will let you resample the image to a different size. Photoshop gives you a limited choice of resampling methods, but as discussed under Q 11 above, there also exists special purpose programs for upsampling (interpolating) to a larger printable size.
To resample to reduce the size of an image is known as downsampling. This is useful if you want to put images on the web or email them to your friends. The number of pixels today's digital cameras produce is much higher than what is normally used on a website, so to put an image on the web, you need to downsample to something like 450 x 300 pixels. The same size is also convenient for sending images as email attachements to family and friends.
If you save the image after changing the resolution with Resample Image checked, the saved image will have a different number of pixels in it than the original.
If you downsample to a lower resolution, the file will be smaller. The data that are removed to make the file smaller are lost and will not be restored even if you at a later point upsample to the original resolution. Always save as a copy if you resample. Leave the original untouched as a digital “negative”.
Acknowledgements: Thanks to Bart van der Wolf, Nostrobino, Roger N. Clark and David J. Taylor for helpful comments and corrections. Any errors that remain are my own.