I've heard that resizing your image from 100% to the desired size is not a good way to resize an image.

Given developer/photographers like Fred Miranda have written tools to help resize your images optimally, how do you optimally resize an image (besides buying FM's cool plugins ;P).

I'm interested in both enlarging and shrinking. I primarily shoot for web, but occasionally will shoot for prints.

  • 3
    It matters whether you are increasing or decreasing the size; can you clarify? Also, for what purpose do you want to resize the image? That matters too.
    – Reid
    Aug 12 '10 at 0:03
  • Are you using Photoshop and, if so, what version? Some of these concepts are based on old algorithms that have since been improved (as of CS2 in Photoshop).
    – Joanne C
    Aug 12 '10 at 0:34

There are a very wide variety of solutions for resizing and sharpening images for use on the web. As preparing images for publication on the web generally involves a reduction in size, you have the benefit of starting with more information than you need. This is always the better position to be in. When reducing an image, use a Bicubic filtering method, rather than a Bilinear, as you'll get better results. (In photoshop, Bicubic or Bicubic Sharper is best.) Sharpening is usually a key factor in producing a quality reduction. Exactly what you are sharpening will sometimes determine which approach you take to sharpening, as not all content is equal. Different approaches work better for different kinds of content.

Sharpening for print is generally more difficult, as if often requires increasing your image size, meaning you start out with less data than is necessary to produce a quality print. It is possible to scale an image up and still maintain high quality, however. When scaling and image up, you can either do it in a single step, or do it in multiple increments. Scaling up incrementally will usually produce better results, as you are iteratively adding new information, which provides each successive resize with more information than if you did it all in one step. Between each resizing step, or perhaps every other step, you might want to sharpen your image to "generate" even more useful data that may be used when resizing. Again, sharpening is a key factor when scaling up, however the process will generally be more complex if you wish to maintain the maximum amount of quality possible. Regardless of what you do, an enlarged image will never contain the same maximum degree of quality as the original.

Image upscaling is an area that is constantly enduring rigorous research, and improvements are made on a fairly regular basis. Many third-party tools exist that employ very advanced algorithms to scale your images (up or down) while maintaining the maximum amount of quality that can be. They often cost a pretty penny, however if you have the cash, they will save to a ton of time (and, depending on the volume of work you do, possibly a lot of money too.)

There are a wide variety of techniques that may be used to resize and sharpen images. Here are links to a few that I have found helpful in the past:


I have found there is no single best way to increase size. I've compared the results from Adobe Photoshop bicubic smoother and Genuine Fractals side by side. Sometimes Adobe does the better job, other times GF. I've tried increasing size in 10% increments, and doing it in one step. Most of the time I see no difference. For me Adobe PS bicubic smoother in one step usually works well. When it doesn't I've rarely found that another method or software can do it significantly better. I think it depends a lot more on how sharp the original appears out of the camera, and careful processing and sharpening.

For reducing size I start with Adobe PS bicubic sharper, and do it in one step. If bicubic sharper makes it look too crispy I go with straight bicubic.


Shrinking an image is easy. Just shrink it, using the bicubic algorithm available in most image processing applications.

Enlarging an image is trickier. Strictly speaking it can't be done, as there is no information to use to determine what the extra pixels you put in should look like. So what you do is to use an algorithm to invent pixels based on the already existing pixels. The bilinear algorithm gives a smoother result that works better for some images, the bicubic algorithm gives a sharper result that works better for other images.

Some say that you get a better result by enlarging in small steps, e.g. repeatedly enlarge by 10% until you reach the desired size.

There are also more advanced algorithms, like the one used by Genuine Fractals that does a better job at guessing what the pixels should look like. It's of course still a guess, you can never get real details back by enlarging an image.

  • I agree that shrinking is easy from, say, ~4000px to ~1000px, but the bicubic is not really the best one when shrinking 20 000px panorama to 1000px. It is noticeably better if you shrink by 10% or just 1000px at a time and apply very light sharpening once or twice in between. Sure, there are even better methods, but this has worked for me those occasional times when I've needed it.
    – Karel
    Aug 12 '10 at 8:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.