After many years of practice, photographers build a sense of intuition when their 35mm negative will be good enough for a certain enlargement. They can tell whether or not the quality will be acceptable for a 120, or even a 4X5. A similar sense of intuition would be valuable in digital imaging, but with things advancing as rapidly as they are - who has the time?
Most people quickly find out that a scanned image that will fit on a single diskette doesn't look very good when enlarged to four feet. We often hear rules of thumb such as, scanned files must be at least 20 megabytes or scanned files must be 100 megabytes or more. In the digital imaging arena, there are few circumstances where general rules of thumb apply.
If you're used to using photography for your digital output, here's one more very rough but useful rule of thumb. Use scanned files with sizes (uncompressed) equal to the film format you would normally use.
35mm negative equals about 20 megabytes of data 120 negative equals about 80 megabytes of data 4X5 negative equals about 200 megabytes of data
These three rules can get you a long way when scanning images, but the rest of this document offers additional information about scanned image size. There are also some invaluable links to WWW sites that will quickly get you up to speed in the wide world of scanning images for large format printing.
Background
There are really two broad classes of digital pictures. One class, commonly called line-art, vector art, scalable, or object art is a snap for enlargement. These graphics are stored in the computer as a collection of connect-the-dot instructions, and moving the dots further apart scales them perfectly. Relatively small file sizes can produce excellent results even at very big enlargements. Typesetting, graphic designs and other spot color designs are commonly handled this way in enlargement, but photographic images are not.
The other class of artwork, known as digital images are made up of a square pattern of points (pixels). The colors recorded at each of these pixels are stored in the computer and this is what makes up the image. It takes a lot of points to cover a large surface, and the larger the image gets, the amount of pixels is drastically increased.
PostScript files commonly contain both line-art and digital images. This is one of the strengths of PostScript. It's a safe bet, though, that if your PostScript file contains digital images, this will be the vast majority of your file size.
Suppose that we scan an image by measuring red, green, and blue (each pixel will be scanned), and we store each of these three colors in one byte of memory. 3 colors at 1 byte per piece equals 3 bytes per every pixel scanned.
Suppose, also, that we are scanning this 8 X 10 image at 300 pixels per square inch (PPI). 300 pixels per square inch means 300 X 300 pixels in every inch, equaling a total of 90,000 pixels. At 3 bytes each, that?s 270,000 bytes of information for every square inch.
The original 8 X 10 has 80 square inches, so we are looking at 80 X 270,000 bytes, or 21,600,000 bytes. Tip: The fact that a 300 pixel-per-inch scan of an 8 X 10 produces a file that is a little over twenty megabytes is a handy number to keep in mind.
Pixels are often referred to as dots when scanning. This is unfortunate because it tends to create confusion between scanned pixels, device dots and halftone dots, which are three entirely different things. Nonetheless, we'll conform to popular usage and refer to PPI as dots-per-inch, or DPI, when scanning.
RGB scanned file (3 bytes per pixel): Bytes = DPI 2 X width X height X 3
This is true where the width and height are in inches, and DPI 2 means squared or to the second power. Note that because DPI is squared, doubling this number will quadruple the size of your file. A scan that produces a 20-megabyte file at 300 dpi will produce an 80-megabyte file at 600 dpi. Likewise, doubling the area being scanned doubles the surface area, and will also increase the file size by a factor of four. This square law behavior is why the size of digital imaging jobs can quickly get out of hand, and why it is valuable to understand just how much resolution you really need. If we were scanning to produce a CMYK file, we'd use a four instead of a three in the above formula. Notice that the effect of this is nowhere near as that of the size and resolution of the picture.
So What Should I Print?
We mentioned above that it is unfortunate that the words dot and pixel are so freely mixed by users of scanners. It is more useful to think of a dot as a physical object, either the spot of ink that a digital printer can place on paper, or the (usually larger) dot that is constructed to form a halftone screen at a particular LPI (line per inch) resolution. A pixel isn't a physical thing at all, it's just a set of numbers that correspond to a color at some location. If we have a lot of these pixels specifying color at very close spacing, we have a high resolution image. If we have fewer pixels, the resolution of our digital image is lower. Hence, the higher the resolution, the more pixels we have, which means larger file size. There are two very important things to keep in mind when scanning images:
1. How many pixels-per-inch (PPI) are appropriate for printing with a given line-per-inch (LPI) or dot-per-inch (DPI) halftone? 2. How many PPI are appropriate for a particular viewing distance/time?
Even in the most demanding cases, there is no point in having 100 times as many pixels as halftone dots. There is, in fact, a limit beyond which we need not go, and this is described by the Nyquist Limit, stated below. As we said in the beginning of this document that general rules of thumb are not usually acceptable in the case of scanning images, you can use the following guidelines to start you in the right direction.
1. Acceptable for most large format: PPI equals half of LPI 2. Pretty good quality: PPI > or = LPI 3. Very good quality: PPI > or = 1.5 times LPI 4. Ultimate quality: PPI > or = to 2 times LPI (Nyquist limit!)
We all want ultimate quality, right? Let's make a 30X40 print with 300-dpi halftoning at the Nyquist limit. Since we're using the Nyquist limit, our scan will need to be 600 PPI and:
File size in bytes = 600 DPI 2 X 30? X 40? X 3 bytes of information = 1,296,000,000 (1.3GB) That's well over a gigabyte file, and most of us know that gigabyte scanned files are pretty rare. They take up too much space and are way too difficult to work with.
Compression and Interpolation
You may be thinking about compression, interpolation and other such techniques. The sad fact is that the LZW compression used in file formats such as Tiff and PostScript isn't very effective on photographic scans. More effective techniques, such as JPEG, just haven't gained wide acceptance in publishing circles. Interpolation can help, mainly by preventing jagged edges, but the jagged edges are replaced by fuzzy edges.
These techniques might get our gigabyte scan down to a couple of hundred megabytes, but the fact is that a great many beautiful 30X40 prints are made from just 20 or 30 megabytes of scan data. One reason is that the Nyquist limit is an absolute limit confining what can be resolved in principle. A human viewer is concerned with the cognitively significant elements of the image.
What happens in the Real World
Although it varies widely between different types of images, the resolution that is perceptually significant is much less than that which is mathematically feasible. In other words, to achieve a great quality print, one does not need to use the highest resolution to scan.
You may never have noticed this, but next time you're in an airport terminal a hospital or even a mall, get within six inches of some of the backlit transparencies. It's a shocking experience to realize just how low-resolution most of these are. And maybe even more shocking is to realize that we may walk by these images and admire them, all the while not noticing the low DPI. Consider a billboard viewed from a freeway. Resolutions of just one or two DPI are commonly accepted.
The real fact is that the choice of appropriate resolution or file size isn't a matter of mathematics and science, but one of art and judgment.
------------------ Bob Burns Bob Burns Signs
1619 Oregon Ave. Prescott, Az 86305 1-520-778-5879
Posts: 2121 | From: Prescott, Arizona, USA | Registered: Nov 1998
| IP: Logged |
Really informative and useful. I used to have a book with the recommended resolutions for different halftone setups, and have missed not having them anymore on many occasions. I've always been overcompensating whenever I needed to just to be sure, which is a pain for all involved and a waste of space. It never failed to amaze me just how low resolution a file could be for general halftones. I've saved this for future use. Thanks.
------------------ Bob Darnell London, Ontario, Canada
If you go down to the part called Scanning Resolution Calculator, it will take you to a great little free calculator. You can even save the htm file to your computer and use the calculator even when not connected to the Internet.
The one key ingredient here is knowing a good lpi value for your printer. On my 300 dpi EDGE, the optimum LPI is 75 though it can print up to 90 on some images. I like 75 because it is easy to double to get the 2x value of 150 dpi. Gerber defaults a placed image to 300 dpi—another convenient value—so an imported image that WAS 150 can be brought back to its exact original height and width by "scaling" the image at 200%.
Anyone using a digital printer should go grab the resolution calculator. Here's the address to just that page:
Printer-friendly view of this topic