Boo's Digital Photography Tutorials.  No.10; Sharpening techniques

If you'd like to see a larger view of any image, please click the thumbnails and then your browser's 'back' button to return here. Because this site is intended as a photography portfolio, the images are rather large and will take a few minutes to download.

Bypass :: Home :: Water :: Loved Ones :: Birds :: Lake District 1 :: Lake District 2 :: Panoramas :: Black & White :: Flora & Macro :: Live Music :: Miscellany :: Wildlife :: Tutorials :: Badgers :: Mail Order Greetings Cards


The black art of digital image sharpening;
personal observations, tips and tricks.

Back to the Tutorials listing.

Do I need to sharpen my images?

Sharpening is very much a subjective matter and seemingly some utopia, digital photographers more than ever, strive towards, as though the sharpess of an image is some barometer of the success in their work. There's far more to a successful image than how sharp it is, but having said that, an otherwise good image can be spoiled by not being sharp enough and thankfully, it is one image feature you can improve considerably after taking the shot.

 [Selective sharpening to the details that actually need it; facial hair, glasses and zip, less so to skin, none on the background.] With a little effort, an average picture can be made into a good one. But it's also true to say, that bad sharpening can equally spoil an image. I personally think sharpening, along with noise reduction and high ISO image management reaps the greatest rewards when a little individual attention is lavished on photographs.

I don't believe there is a one stop solution to achieving good sharpness and each image will need something different, determined by the subject matter, starting quality of the image, the format in which it was recorded and the camera settings, end use and the size of the end product. An image file of a child's face to be prepared for printing at 20" x 16" will need a very different solution from an architectual subject to be viewed on screen at 800 pixels wide.

 [Enhance creative DOF by only sharpening foreground details.]  The problem is, there just isn't one solution to fit all eventualities, the best sharpening results come with practice and to a very large extent, are dependent on personal taste. There are many ways to sharpen an image, different software, plug-ins and filters and we all probably have our favourite techniques and stick to a selection of them. I doubt there's any wrong or right and for me, technique is constantly under review and I graduate through different approaches as I go along, constantly trying and settling on better ways of doing things. It's also true to say that I perhaps sharpen less than I used to, my tastes have changed and I also wish I had the luxury of time to revisit earlier work and improve them with the benefit of increasing experience.

This page is some of my personal observations, favourite techniques and ideas based on lengthy retouching experience. I'm not suggesting it's the right or best way to do things, but very much my personal approach.

Where on earth should I start then?

Conventional wisdom says that you should only sharpen an image twice in its workflow - you can do more that one pass of a sharpening routine at the same time, but only at two separate stages in the workflow. If you're working with JPEGs saved in the camera and have some sharpening enabled in your camera settings, that becomes your first set of sharpening, so you shouldn't need to apply any more - and if you do, do so gently. If you re-sample that image, say to a smaller size for web publishing, you'd then sharpen again - modestly - at the resampled size, just to remove the softness that is introduced by the re-sampling process. As such, your out of camera JPEG image has already been sharpened - you don't need to do so again.

 [This 1600ISO shot wasn't sharpened at all until ready for publication at this size, then only the subject was sharpened, not the background areas.]  I routinely work in RAW and I develop the images in C1 with sharpening disabled on output. That way I can treat each image individually dependent on what the image needs and what I'm going to do with it. It has been my experience that images prepared for printing can withstand a great deal more sharpening than images prepared for on-screen viewing and I usually treat images needed for both in two separate workflows.

If an image needs a lot of retoucing work, I develop it as a 16 bit TIFF and do such work at full resolution and keep that file saved as a working master - at this stage I have still done no sharpening. I'd then prepare a separate final copy of the file for each intended end use. It usually only takes a few extra minutes, but I have found the results very well worth it - I get it cropped exactly where I want and the right degree of sharpening for the print size being ordered.

A separate approach for print and screen - why?

 [A 1:1 pixel crop from a file prepared for print.  At this size, it looks over-sharpened on screen.  ] It might sound like tedious levels of duplicate work, but I personally don't print lots of images, most are for web publishing and print and screen need a different file for optimum results. Most of my printing is done at a lab and after some experience of ordering I know what the file needs to look like when it leaves me to get a print I'll be happy with. If for example, I want an 8" x 10" print, I crop the image to the right proportions and resample to something like 2400 x 3000 pixels which is 300 pixels of image per linear inch of paper, a good resolution for maximum detail, although less will still give good results. I then sharpen this file, at this size, for printing.

 [A screen grab of the image above, prepared for printing, but viewed on screen at 40% magnification, which gives a good indication of the finished print appearance. ] I've got into the habit of viewing print-ready files at about 40% to 60% in a particular editor with a good rendering engine and I can tell if it's sharpened about the right amount from how it looks on screen at that point - see right. It will almost certainly look too sharp for screen viewing when viewed 1:1 pixels (above right), but prints can usually stand much more sharpening than looks good on screen.

Keep in mind, when you print, you're effectively resampling the image smaller - so view it on screen at a size more like the finished print size than 1:1 pixels, to get a better indication of the anticipated result. If I want another size print from the same image, I repeat the crop and resample routine for each size and save and name the files accordingly, although I perhaps wouldn't bother for similar sizes.

[ The same image sharpened and finished for screen viewing. ]In my prepapration of images for screen viewing, I have several approaches, depending on the quality and nature of the image itself. High ISO images I tend to resample without sharpening at all, then only sharpen the details in the image and the focus of attention where it actually needs it at the published size, leaving much of the image alone and totally un-sharpened, which tends to help the noise situation enormously and is often enough in itself to avoid the need for further noise reduction procedures.

Images that are otherwise in good shape, low ISO and have detail or texture evenly over the frame, such as landscapes, I often sharpen very hard at full size, even more so where the size change will be considerable, then resample to the desired size and then apply a very small amount of 'polishing' sharpening to counter any softness introduced by the resampling process and perhaps small localised amounts to further bring out details. But very many images don't fit easily into this global approach, for either print or screen and many subjects benefit from a more individual and tailored approach.

An individual approach; lavish some attention on those images

 [A before and after screenshot to show how careful sharpening can be applied just where required.]  As with many post-processing techniques, sharpening is one where mastering the use of layers and masks is very beneficial, it gives you enormous control and flexibility and once you become confident and proficient, it's a quick and efficient way to apply sharpening, along with other adjustments, selectively and to varying degrees across an image. Many images don't need sharpening applying globally and woking on layers allows you to apply as much as is needed, where it's needed and areas that might be harmed by sharpening can be left alone, or minimal levels applied.

Where images warrant this level of attention, I usually work at the published size for more accuracy and there are two main ways in which I use layers to apply sharpening. The simplest is to apply sharpening to a copy of your open image at a level that perhaps looks a little too much. Then copy the original unsharpened image as a new layer above the sharpened one and adjust the transparency of the uppermost layer so that it blends with the sharpened copy underneath - simply tweak the transparency of the top layer until you like how it looks. It's very simple, but adds much more flexibility and you can get it just how you want it. Depending on the editor and tools you use, you can alternatively just duplicate the image layer and sharpen one copy, then adjust transparency.

[The red area shows the extent of the sharpening mask, avoiding the background and his adolescent skin.  ]  But the real power comes when you bring masks into the mix. If you repeat the same process outlined above, so that you have two versions of your image on top of each other, one sharpened and the original above, but now add a mask layer too. A mask allows you to paint on, or otherwise add, areas of transparency to allow the sharpened image to show through from beneath - you basically paint holes of transparency into the uppermost layer. The really flexible aspect comes from the nature of the mask - you can create it with up to 256 levels of grey; black being fully transparent and white remaining opaque. You can therefore allow the sharpened copy from beneath to show through just where and by how much you want.

 [The finished image, selectively sharpened and softened to taste. ] My own preferred technique for creating such a mask, is to simply paint the areas I want with a brush tool, using different greys, until I like how it looks, by constantly removing the visual indication of the mask to see the results. When I'm satisfied with the results, I apply some gaussian blur to the mask to soften its edges and blend the tones, then merge the layers. It is nowhere near as complicated as it might sound and you don't even need to be that careful in your painting, if you don't like any particular bit, just paint with a lighter colour, or to totally remove it, paint it over with white. If you really mess up, just delete the mask layer and start again.

In the example image above, I sharpened one copy with a large radius to improve contrast on the dark suit fabric and a pass of a smaller radius to sharpen details. The unsharpened image was masked to reveal sharpening to differing degrees around his front hair, glasses and clothing (as shown in the mask image above), but not his rather pimple-ridden adolescent skin. The background was also softened in a separate process to take it further back and to prevent it distracting from the subject. A cable was cloned out by his shoulder as this caught the eye too much.

If you don't want to use layers, lassoo it instead!

If you don't feel confident using, or want to fiddle with, layers and masks, you can still selectively sharpen, just use a freehand selection tool with a feathered edge so you don't get harsh divisions and select/lassoo the areas you want to work with. This still allows you to apply just as much or as little sharpening as you want - just where you want it. Identify areas you want to work on, select them and sharpen to taste, then select the next area, where a totally different type/level of sharpening can be applied.

 [The only area of this 800ISO shot substantially sharpened was the nose, eye and whisker area.  ]  With portrait shots, which are often cited as problematic to people, I'd sharpen areas like eyes much harder, leave backgrounds totally alone and moderate amounts of sharpening to detail areas like clothing and jewellery. Depending on the age of your subject, skin usually is best just left alone, sharpening simply isn't necessary on young skin, tends to enhance blemishes on teenagers and more mature subjects would find softer skin more flattering. In fact on mature subjects, I'd probably selectively soften skin in the same way I would sharpen other details.

In the 800ISO cat shot shown above right, the out of focus background areas were left totally unsharpened to minimise noise, the fur in focus was given a pass of USM with a small radius and the eyes, nose and whisker area sharpened rather more, as this is the focus of attention. Animal fur and human hair can quickly get very ugly, with blotchy artefacts, if oversharpened and as long as the eyes are sharp, it simply doesn't need it, so don't be tempted to overdo it. Things that nature made soft, like skin and hair, are probably best left that way in photographs too.

Let's crunch some numbers

 [In this 800ISO open tulip shot, I only sharpened the centre flower details to minimise noise elsewhere. ] The one thing I haven't touched on yet, is how I actually sharpen the images and I suspect this is what most people want to know. There are many ways to do this, but my favourite, despite flirting with other techniques periodcially, is still to use Unsharp Mask (USM).

You have a huge amount of control with the three factors and if used creatively as outlined in the methods above and if you understand the implcations of changing and mixing those settings, it's still a very effective way to sharpen. Most proprietary sharpening tools use USM, the clever bit is in the settings, combinations of settings and blending and layer modes they automate for you. I just happen to like doing that for myself.

I have this mental idea that sharpening is rather like polishing something. You start with a course layer of abrasion to take off the roughness, gradually using finer techniques, with a final finish with a very fine polish worked hard to bring up the shine. To this end, I routinely run several passes of USM with gradually dimishing amounts. I often sharpen quite hard at full resolution for web images and final size for print files, then apply a much finer pass of USM at the smaller published size.

[ See how a larger radius causes halos and artefacts around edges.]  Perhaps the safest rule to keep in mind, is that a large amount of a small radius is most likely to give pleasing results and least likely to introduce sharpening artefacts and perhaps the safest approach to adopt. As you increase the radius size, the sharpening looks progressively harsher and bright halos appear around edges, especially dark ones against light backgrounds. The screen grabs above right, of a hand written price ticket in a market shot, shown at 300% on screen, shows the nasties that occur as the radius is increased, even when the amount applied is less.

The final sharpen pass I do after resampling an image is the 0.3/500/0 setting, often run twice - that's just enough to brighten any resampling softness without looking sharpened. A radius of 0.3 is about the radius size limit you can sharpen without introducing any nasties, so it's a powerful figure to keep in mind.

Experience would suggest that about the maximum radius that's routinely useful for general work is that of 0.8. If applied in a small amount, say 50, it's good for brightening eyes and harsh features like architecture, metalwork, jewellery etc. We probably all have some favoured combinations of numbers we use for different subjects and I certainly have scripts set up for my regular combinations. 0.8/50/0 is good for small areas that require attention, as described. 0.6/100/0 I tend to use for global areas in textured subjects for enhancing detail, such as landscape shots. 0.4/200/0 is a good global and gentle amount for general use, lowering the amount to 100 where less is needed and the 0.3/500/0 I use for my final polishing.

Contrast enhancement using USM for perceived sharpness

 [A before and after screen shot of the sharpening applied to an 800ISO image.]  There is one setting that I perhaps use most often and is a very powerful way to enhance images of just about any nature, giving images the 'punch' or 'pop' photographers regularly ask how to achieve - and it rather flies in the face of what I just told you - use a very large radius!

A large radius, in excess of 10, will have the effect of increasing local contrast and is very good for removing the haziness of some images and boosting the overall impact of image details. It increases the visual perception of sharpness without actually sharpening image details, so is especially good for high ISO images by making them appear sharper without having any impact on noise pixels. My starting point for this technique is radius 20 with amount 20.

The figures can be increased quite high from there - a little bit of trial and error might be necessary, but most images can be improved by the application of a large radius pass of USM. Radius 10 with amount 10 will add a very subtle increase in contrast and texture and yet won't have any negative impact at pixel level.

 [The finished and selectively sharpened 800ISO available light shot.]  The indoor market shot right was worked with selective sharpening using layers and masks, as it was taken at 800ISO and this level of care can really have a positive impact the final quality of high ISO images and is worth the effort. I duplicated the image in my software at the resampled size shown and sharpened one copy with a large radius pass of USM (20/20/0), followed by two passes to actually sharpen details, one at 0.6/100/0 and another at 0.4/200/0. These levels might be too much on many images, but the detail and texture in this photograph could stand it.

I then applied a tiny radius (0.2) of Gaussian blur to the unsharpened version to slightly soften the background noise in the out of focus areas, also helping with the depth of field, probably only worth doing due to the simplicty of the image shapes in this instance.

I pasted the unsharpened version as a new layer above the sharpened one and painted a fully transparent mask over the foreground items, easily done with such a simple shape. I painted just short of the subject edges, applied a gaussian blur of radius 8 and merged the layers. The image is a very simple merge between a sharpened foreground image and a very slightly softened background, which took much less time to do than to explain!

If you found it useful . . .

Boo's digital photography tutorials are provided free of charge for fellow photographers, but if you found them useful and would like to make a small donation to the upkeep of the site, you can do so using PayPal by clicking the button to the right. Thank you.



Back to the Tutorials menu

© 'Boo. 2002 - 2005. All rights reserved. boo@zen.co.uk ~ Site content must not be used elsewhere without consent.