Sharpening
by Nick Wilcox-Brown
Written for Canon Professional Network
Contrary to popular belief, sharpening is not a filter devised to rescue out of focus or ‘soft’ images. Instead, it is the process by which detail in an image is revealed and enhanced.
Sharpening brings images to life
Digital sharpening has the effect of bringing an image into focus. Many think of the effect as magical when they see it for the first time; the effect is to show image detail that did not appear to be present before.
What actually happens is that any edges within the image are enhanced, adding contrast and therefore apparent definition, or sharpness, to the image.
Next only to colour management, sharpening has proved to be one of the most elusive concepts for many photographers to grasp; a situation made worse by the fact that many picture desks and clients have even less idea than photographers themselves.
Evolution
Sharpening is based on a traditional process from the days of film and darkrooms. An out of focus negative was made and sandwiched with the in-focus original. By creating a halo effect on the edges within the focused negative, the image appeared to have an enhanced resolution.
Once film scanners became commonplace, sharpening became an optical process, applied within the CMYK drum scanner. In the magazine and print world, images were mocked up in layout which was sent to the printers. The repro [reproduction] house, usually part of the print company, took the transparencies (or prints) and scanned them to a specified output size as the final stage before the document went to print. Sharpening settings were applied specifically for the size of the image in the layout by highly skilled scanner operators and the scans were for a single use only. If the image was to be re-used elsewhere, it would have to be re-scanned and sharpened as appropriate.
With the advent of digital capture and the effective death of scanning, many of those skills and practices have been lost in the space of less than 10 years.
Digital workflow necessitates capturing a fixed size original and re-purposing it for multiple uses. The problem inherent with this system is that the original file produced by most digital SLRs is far from optimal sharpness and has a tendency to look soft and slightly de-focused. This is due to the effects of the camera’s anti-aliasing filter designed to remove moiré patterns. Combine with this, the losses from the print process where the combination of four pixels into one print dot produces a marked sharpness reduction, and a solution is clearly needed.
The answer to this is surely to add our own sharpening? The answer has to be ‘Yes’, but without specific information about size of reproduction or the usage, it is not easy to make accurate decisions on sharpening settings.
Good sharpening makes an image, bad sharpening will destroy it.
A new workflow
To cope with the changing way that images are captured and used, a new sharpening workflow is evolving. The rules are established, but not yet fixed, and at this moment, the practitioners tend to be photographers because the rest of the industry has yet to catch up.
Instead of the traditional way of working, with sharpening being the final step before print, it is now applied at two or more stages, depending on the way that the image is being used.
Capture sharpening
Performed post capture, before the client first sees the image. Typically capture sharpening is a straightforward process of enhancement to best present the image on screen or a contact sheet.
Output sharpening
Performed when the image is prepared for print. This may be a single operation or involve localised enhancement of areas within the image. Much is dependent on the image and the requirements and quality of the printing process. Sophisticated enhancements are unlikely to bring benefits in newsprint, but will undoubtedly do so in an art book or exhibition print.
In addition, we should make mention of Screen Sharpening, specifically for web use. Although similar to Capture Sharpening, the images will be at a much lower resolution (72ppi or similar) and will need separate treatment.
How to perform sharpening?
Photoshop has available no fewer than 5 sharpening filters. Of these, only two serve a useful purpose. The older, but utterly reliable Unsharp Mask (USM) or the new, but more fickle Smart sharpen.
Smart sharpen is a sophisticated but complex filter and deserves a detailed explanation of its own. For the purposes of this piece, we’ll concentrate on the older, but very capable USM.
Unsharp Mask
Unsharp Masking, otherwise abbreviated to USM, is used in Photoshop and most other general photographic applications. Despite the recent appearance of other forms of sharpening, USM is still the favoured choice of many and a trusted tool that is both straightforward to to use and easily controllable.
Correctly sharpened, the cranes retain detail, but none of the unpleasant haloes exhibited in the preview window.
Unsharp Mask works by enhancing the contrast on either side of edges. The preview window of the tool shows a deliberately exaggerated view of this, with oversize light contours around the dark edges of the image. When used in the way shown above, the result is ugly, but used in a carefully controlled manner, the impression is of added definition and increased sharpness. Images seem to ‘pop’ and come to life.
Using USM
The first sight of the controls for unsharp masking make the tool look daunting. With a little practice however, all becomes straightforward and good results are easy to achieve. The watchword here is caution.
Amount:
Increases the sharpening effect as the percentage is raised. The higher the amount, the larger the sharpening halo will be.
Radius:
Radius determines the character of the sharpening. A small radius creates small sharpening haloes and brings out fine detail in an image. A large radius creates a much wider halo and far more obvious sharpening effects. The scale indicates the radius of surrounding pixels included in the calculations.
Threshold
Threshold is easiest described as the protection factor. Threshold sets the brightness difference there needs to be between pixels before they are considered to be an edge and sharpening is applied. For maximum effect on low noise images, use a zero or low threshold.
In practice
Screen or Web Sharpening:
Screen files are always at low resolution, so great care should be taken to ensure that images look their best. A good starting point is Radius 0.7, with a percentage of 50%. Threshold 1.
Depending on the image, the percentage may be increased or for very fine detail, the radius reduced to 0.5 and the percentage taken to 75 or 100%. If noise become obvious in skies, for instance, increase the threshold to 2 or 3.
Capture sharpening:
A lower radius with a higher percentage is generally accepted to give the best results. Typically a radius of 0.3 - 0.5 is ideal, with percentages in the order of 200 to 300% and a threshold of 2-4.
Practical differences between Capture Sharpening and Output Sharpening.
To left, the capture sharpen was done with settings of 300% / 0.3 / 0. To right, the output sharpen was done with 350% / 0.9 /3
Specifics are hard to recommend due to the differences in output from different cameras, even within the Canon range. The nature of the source images also has a bearing. Some may prefer to work with camera derived JPEG files, with a hint of sharpening, whilst others will prefer to work with RAW derived images that have no previous sharpening.
Images with more fine detail may benefit from settings at the lower end of the recommendations, whereas noisy, or high ISO images may be better treated with a much higher threshold and radius settings, but a lower amount.
Output sharpening:
For single stage output sharpening, the principle is generally to use a larger radius to give a more overall image sharpen and to compensate for image re-sizing that may have been done since the capture sharpening.
It is normal to set the radius first as this is determined by the intended use and the image itself. For a detailed image to be used in a book or glossy magazine where close viewing is likely, the radius should be in the region of 0.7 to 1 pixel, with an amount of around 350%. For newsprint, this is more likely to be 0.9 - 1.2 pixels because of the coarser printing dot.
Images used for larger display use will require a radius in the region of 1.5 - 3.5 pixels or more. This is because the resolution of the printer will be much lower and the viewing distance is likely to be significantly greater. Thresholds for these settings have to be judged on an image by image basis depending on content and image noise.
More advanced techniques for sharpening images abound. A popular myth exists that converting to LAB colour mode, then sharpening the Lightness channel before finally converting back to RGB colour mode will help reduce image noise. Whilst there is some truth in this, the often overlooked issue is that for an 8-bit image (such as a JPEG), the penalty is the loss of around 10% of the levels information due to quantisation (number rounding) errors. It is actually better to sharpen a duplicate layer in luminosity mode and then flatten the image afterwards.
There are numerous ways to work with sharpening with layers, layer masks and other advanced techniques. It is definitely worth a Google for other ways of working.
Care needs to be taken when sharpening in shadow areas or noise may be bought out in the image. A threshold of 6 was used here in combination with a larger radius for coarser detail sharpening
If this is all proving to be too much, there are two excellent products that can help to bring some automation to your sharpening:
Software
Photokit Sharpener has been developed by a small group of Photoshop gurus and does an excellent job at sharpening images.
<http://www.pixelgenius.com/photokit/>
Nik Sharpener Pro is a must-have plugin that is quick and reliable and can be set up as a Photoshop action to provide reliable batch processing on large numbers of images.
<http://www.niksoftware.com/sharpenerpro>
Copyright © Nick Wilcox-Brown / Canon Professional Network 2007.