Monday, February 20, 2012

Image resolution for digital photography

Creating the sharpest images possible is a holy grail for many photographers.  It's easy to get the idea that image sharpness comes only -- and always -- from bigger, more expensive gear.  This isn't necessarily true, but even if it were, you may not need better equipment at all depending on your application and processing workflow.  Read on for info on how good is good enough when it comes to image resolution.

Making Sharp Prints

At what image size must you record your photos in order to make sharp prints? That depends on how large the image will be displayed and from what distance it will be viewed. In order for an image to appear sharp, individual pixels needs to be displayed small enough that they appear to be points rather than small disks. For most humans viewing an image from about 12" away, the maximum size of a "sharp point" is about 250 microns (1/4mm or 1/100"). Doubling or halving the viewing distance means doubling or halving the maximum pixel size to retain sharpness. Small photos, like Christmas cards, are often viewed from a much closer range, requiring a considerably smaller point size (i.e., higher resolution).

For large prints hanging in the foyer which are rarely viewed from closer than 12", this means that a resolution of 100 dpi (dots per inch) is probably good enough, and 150 dpi or 200 dpi is more than enough. For a 16x24" print, the printer then requires an image size of 1600x2400 (3.8 MP) to maintain sharpness. Increasing the resolution to 150 dpi or the print size to 24x36" requires an image of 2400x3600 (8.6 MP). While 3.8 MP may seem very forgiving given today's sensor sizes, remember that this allows no room for cropping the image, and 100 dpi provides barely acceptable sharpness when printed. If you plan to crop your image, remember to take the new image dimensions into account when calculating your maximum printable size.

Note that in order to make a noticeable improvement in resolution (41%), you have to double the number of pixels on your camera's sensor. If a 4 MP camera can produce a 100 dpi print, then an 8 MP camera can only produce a 140 dpi print of the same size, and a 16 MP camera is required to produce a 200 dpi print.

Here's another tip regarding the printing process: most modern photo printers operate at resolutions in the 1200-4800 dpi range. These are far higher pixel dimensions than your camera can produce at anything but a wallet size. In order to feed the printer the pixels it's expecting, software somewhere along the way will fill in the extra spaces between the pixels you provide with similar pixels so that what hits the paper resembles what's on your screen. There are different places where this replication could take place. You could do it yourself in Photoshop before you print the image, or it could happen in Photoshop's "print" back end, or in the printer drivers. The latter stages are definitely preferred, as they will produce the best results possible. Attempting to resize the image yourself before printing by modifying the resolution can do no better than letting the drivers handle it automatically, and if you don't know what you're doing, you're more likely to make it look much worse. Just let the printer drivers do their job and don't try to interfere by forcing the resolution to something larger than what you've got.

If you do need to resize an image for whatever reason, you will likely be given the choice of several different methods for filling in the spaces between the original pixels, known as "interpolation algorithms." "Pixel replication" or "nearest neighbor interpolation" simply duplicates the original pixels, and tends to result in blocky images. "Linear" interpolation gradually changes the RGB values between one original pixel and the next. This is fast, and produces decent results. "Bi-cubic" interpolation does a better job with these gradual changes.  Some software will have the option of "Lanczos" interpolation. It's the slowest method, but produces the best looking results when enlarging an image. Significantly enlarging an image like this will still produce a blurry image, but it's the best you can hope for, and will look fine from a distance.  Check out the following examples of several different interpolation algorithms on an image enlarged 40X.  Click the images for larger versions.

Nearest Neighbor Interpolation (None)
Linear Interpolation
Bi-cubic Interpolation
Sinc (Lanczos3) Interpolation

Excited to start printing your photos?
Shop for photo printers at


Making Sharp Images

Keep in mind, too, that just because your camera has enough resolution to produce a sharp print doesn't mean that the photos you take will look sharp. Unless you employ good techniques to reduce blur, a photo that looks fine on your camera's tiny LCD or even when scaled to fit your computer monitor may still be quite blurry when viewed at 100% magnification. That blur will be even more evident in a large print. Fortunately, there are a number of ways to reduce image blur (or to create it, if you're a "glass half empty" kind of person).

First, when recording JPEG images, always set your camera and software to the highest quality settings possible. These may be labelled as "super-fine" on your camera or "95-100%" in your computer software. JPEG is a lossy compression algorithm, which means that every time you edit and save an image, you loose detail. This typically shows up as noise or blockiness. The lower the quality at which you encode the JPEG, the more information you'll lose. Lower quality images require less space on disk, which makes them ideal for sending through email or uploading to a web page as long as you never expect to print them. If you plan on making multiple edits to an image, it's best to save interim copies of the image in a lossless file format such as TIFF. TIFF's "LZW" compression algorithm (which is not always used by default) can provide smaller file sizes that aren't much larger than those of high-quality JPEGs, but without any loss in image quality like you'd see in a JPEG. Incidentally, the old GIF file format also uses this same LZW compression algorithm. The new PNG format, which is gaining popularity for web images, is also lossless, but uses a different algorithm. You can edit an LZW-compressed TIFF image as often as you like with no degradation in image quality. TIFF is an old, well-established file format that any professional printer and most consumer software will be able to handle just fine. Canon's CR2 raw format is, in fact, just a TIFF file with a couple of extensions.

Another way to improve image sharpness is to pay attention to your shutter speed and make sure it's fast enough to eliminate any visible shake from unsteady hands. For an average person using a camera with no image stabilization, the slowest shutter speed at which you can reliably produce sharp, handheld images is 1 over your focal length (1/50s for a 50mm lens, 1/300s for a 300mm lens). Note that the focal length mentioned is the effective 35mm "full frame" focal length. Most low-end digital SLR's (those using APS-C or "crop" sensors) must multiply the lens' true focal length by 1.5 or 1.6 to get the effective focal length. Point & shoot cameras have considerably smaller sensors, and must thereby multiply the printed focal length by 5 or 6 times to get the 35mm equivalent. Check your owners manual for the actual number. Image stabilization will generally allow you to use a shutter speed that's 4-8 times slower than the above number before you start getting blur from camera shake.

One final method for improving image sharpness is to shoot at a medium aperture, typically 2-3 stops below "wide open" for a given lens. All lenses suffer from distortion known as "spherical aberration" at wide open (numerically small) apertures, while a different form of distortion known as "diffraction" starts to take over at small (numerically large) apertures. Crop sensor SLR's can get diffraction-free shots down to f/11 before an 8x12" print starts to get blurry. Point & shoot cameras can often only go to f/4 or f/5.6. (This is why landscape photographers prefer cameras with the largest sensors they can afford.) Bob Atkins has written several good articles with more information on this, like this one.

Oh, and never use digital zoom if your camera has it, because it merely does a crop and resize, but usually using a fast (poor quality) interpolation algorithm. Just zoom as far as you can optically and then crop if necessary on your computer.

Also remember that increasing the pixel count on your camera may give you sharper images in ideal lighting conditions, but it can have a detrimental effect on image quality in sub-optimal light.  Read more about that in an earlier post.

Hopefully this gives you a few pointers on how to make the best prints possible.  Depending on your intended application, you don't necessarily need the most expensive gear to get acceptable results.  If you've got any questions or other thoughts on this topic, please speak up in the comments below.

No comments:

Post a Comment

Please leave your comment below. Comments are moderated, so don't be alarmed if your note doesn't appear immediately. Also, please don't use my blog to advertise your own web site unless it's related to the discussion at hand.