Digital images need sharpening to overcome the effects of interpolation and anti-aliasing filters. Additional sharpening is often needed for creative effect, and a third round of sharpening is often needed for optimal reproduction, whether that is on screen or in print.
Sharpening of digital image files is one of the most important aspects of image quality and arguably one of the least understood. Sharpening brings out detail and gives an image presence, but not all images should be sharpened the same and even areas within an image often need a specific sharpening treatment. When it comes to sharpening, there are many factors that need to be considered.
|Figure 1 No sharpening (left), correct sharpening for screen (center, using PhotoKit), and over sharpening.|
The most fundamental sharpening issue is to understand the concept that it is best handled as a three-step process. Digital images have varying degrees of sharpness when cameras or scanners create them. Overcoming the slight blurring effects of filters or sensor design is termed "capture" sharpening.
Images have different characteristics and often need more sharpening in some areas than in other areas. This is termed "creative" or "process" sharpening.
Once a digital image is a finished high resolution file, perhaps a master file, it will need a third and final sharpening which needs to be tailored to the type of output. There are four basic outputs requiring different sharpening amounts:
- Continuous tone prints
- Ink-jet prints
- Commercial printing
The need for capture sharpening depends on the sensor technology, the presence of blur filters, lens quality, and f/stop settings. In general, the best methodology is to accept that any new camera model is likely to have different requirements from any camera model you are currently using. Only testing can tell you how much capture sharpening should be applied. Often, researching forums or camera reports can give you a good idea of starting points.
- Does the camera use any type of blur filter over the sensor? Check the specs. Most DSLR cameras use an anti-aliasing or anti-moiré blur filter over the sensor. Many medium format backs do not use blur filters; multi-shot, scanning backs and Foveon X3 sensors do not need blur filters. The presence and strength of a blur filter will determine how much "capture" sharpening may need to be applied.
- Does the camera use a color filter array? Cameras with color filter arrays, such as the Bayer filter mosaic array, create images with some slight blurring due to the color interpolation that creates all three RGB colors from each of the single red, green, or blue color pixels to create a full color image.
- Are the original captures JPEG or raw? JPEG capture "bakes in" any in-camera sharpening settings. For raw capture, only the preview JPEG is affected by in-camera sharpening settings. With raw capture, a high in-camera sharpening setting can give the back of the camera image better definition without affecting the raw file.
|Figure 2 In-camera sharpening on and in-camera sharpening off|
- Does the camera sensor detect and remove digital noise, and is that processing reflected in the raw files? Many DSLR cameras have noise removal algorithms built into the sensor circuitry. This can affect the need for sharpening in two ways. Noise removal can make images softer, so they may need more sharpening, but if noise is not removed, sharpening can accentuate noise.
- What ISO settings were used, and how does that affect noise in the image? In general, higher ISO settings result in noisier image files. This varies considerably depending on sensor size, design and, in the case of raw files, the characteristics of the raw processing software used.
- Were the exposures made at small f/stops such as f/16 or smaller, and how much diffraction effect has occurred? Lens diffraction has always been a factor in image sharpness. Digital sensors add a new twist with the size of the photo-receptor, as does the presence and efficiency of micro-lenses affecting diffraction. As a general rule, high photo-site densities mean that images begin to lose detail at small lens openings. This diffraction blur can be overcome by judicious sharpening.
Some images have high edge frequency, some have low edge frequency, and some are a combination. Edge frequency can be described as how many alternating patterns of strong lights and darks appear across the image. For instance, a tree with many fine branches is a high edge frequency image, while a large smooth object (a pumpkin is often used as an example), is a low edge frequency image. A tight portrait is a good example of a combination image. The areas of skin are low edge frequency, but the hair and eyes are high edge frequency. Portraits are among the trickiest of images to sharpen properly.
|Figure 3 A low frequency image (left) and a high frequency image (right).|
Image "look" or "presence"
What sharpening "look" do you want to give an image? Here is the subjective aspect of sharpening. We can't tell you what that look should be. You'll need to arrive at that determination through testing. Our goal is to help you understand the parameters, and to understand that sharpening for the output is the way to preserve the look you choose.
What is the final output size? The final output size (and output resolution) plays a big role in the required amount of output sharpening.
What is the final output media: screen, offset, ink-jet, or continuous tone? Different media have different sharpening requirements. Computer monitors have only about a third of the resolution of most print output. Lower resolution requires lower sharpening settings, especially radius settings. Since print has higher resolution, and another feature called "dot-gain" where the ink spreads slightly when it hits the paper, print output requires higher sharpening settings. The difference between screen and print media makes judging proper sharpening for print difficult to evaluate on a computer screen.
Will a multiple exposure blending technique be used in post production? When multiple identical digital images are aligned and blended, digital noise can be largely eliminated. This allows for more aggressive image sharpening of the final blended result.
All image files, regardless of reproduction method, require output sharpening. This targeted output sharpening occurs when the image is at the final size for the specific display or printing device and substrate. Sometimes, the exact reproduction size is the only piece of missing information preventing delivery of a completely finished image file. While it is possible to sharpen delivery files “generically”, any resizing done after this will resample the pixels and result in some loss of sharpness. As mentioned previously, a work-around is to deliver sets of files at different sizes, but this does add to the workload. However, use of Photoshop actions can automate and speed up this process. There are two fundamental ways to sharpen digital images: sharpening the pixels that make up an image, and edge sharpening, which is sharpening the edges of shapes within an image.
We believe that while capture sharpening can be done inside Camera Raw or Lightroom, output sharpening is best done within Photoshop. Use of pixel sharpeners such as Unsharp Mask, Smart Sharpen filters, or sharpening plug-ins such as PhotoKit Sharpener, Nik Sharpener Pro, and Focal Blade can do a good job. The plug-ins are easier to use than Unsharp mask or Smart Sharpen because they have built-in parameters for different sizes and outputs.
However, we prefer an edge sharpening technique for most output. Edge sharpening avoids emphasizing noise or other artifacts. In addition, edge sharpening holds up better if an image is resized slightly — which is often an unavoidable occurrence when image files are put into page layouts. The best technique for edge sharpening is to use one of the many variations of combining the Photoshop high-pass filter with an overlay – hard light, soft light, or linear light-blending mode. A nice advantage of this method is that it requires a duplicate layer, which makes the process non-destructive and infinitely adjustable by tweaking the opacity of the high pass layer. This technique also allows for adding a layer mask to prevent some areas from being sharpened or to minimize selective sharpening.
Good resource for learning about sharpening
We found one of the most succinct and best explanations of the sharpening tools in Camera Raw and Lightroom in The Ultimate Workshop by Schewe and Evening.
How to judge sharpening for output (it's tricky)
When judging sharpening for print, the image should be viewed at 50% or even 25% (if is a very large image), and not at 100%. Viewing at 50% gives a much better approximation of the actual effect of the sharpening whereas the 100% view will be largely misleading. Appropriate sharpness is definitely a subjective decision. Our advice is to try many techniques until you find one that gives good results and is repeatable. Keep a record of what you like best so you do not have to recreate this part of the wheel each time. Remember that different output devices as well as different substrates may each require very different approaches and levels of output sharpening.
Raw processing applications, otherwise known as parametric image editors (PIEware), almost always have sharpening functions built in. These have become increasingly sophisticated, graduating from simple sliders to incorporating not just the amount of sharpening, but other parameters such as radius, detail and masking.
|Figure 4 ACR sharpening menu|
We find that testing is required for different images with varying edge frequencies to understand exactly how these ACR controls work. Our opinion is that the settings should be selected to provide good capture sharpening for your particular camera. We find that the default settings for radius, detail and masking are a pretty good compromise, and that you may want more or less, depending on your camera. If you use other PIEware, you'll find different controls and different effects. One issue of concern is that some PIEware sharpens images even when the sharpen slider is set to zero. This is due to some sharpening being built into the demosaicing algorithms.
It is our opinion that PIEware sharpening works best for capture sharpening. Images should not be sharpened "for effect" if additional image editing in Photoshop is planned. Sharpening when applied to pixels, as with files exported to Photoshop, is permanent and somewhat destructive. Photoshop's tools, especially layers and layer masking, are much better tools for creative sharpening and output sharpening.
|Figure 5 ACR has an output sharpening function, which it shares with Lightroom, but this should only be used if you are printing directly from Lightroom, or using Photoshop as a printing pass-through from ACR.|
Photoshop tools: Sharpen, Sharpen More, Unsharp Masking, Smart Sharpen, High Pass Filter
As has been said many times by many people, the Sharpen and Sharpen More tools are blunt instruments with no adjustabililty, and serious workers should avoided them. Unsharp Mask is generally accepted as the best tool, having enough adjustability to achieve nearly all sharpening goals. With the introduction of the Smart Sharpen tool in CS2, you might ask, why not just choose Smart Sharpen and be done with it? It's the smartest option right? Well, not exactly. Maybe Adobe means you have to be extra smart to use it since it has so many levels of control. The primary advantage to Smart Sharpen is that it has a function to overcome motion blur, which is not available in other Photoshop sharpen tools. The added complexity of Smart Sharpen is meant to address sharpening Shadows and Highlights differently, a worthy goal arguably better addressed with regular Unsharp Mask in conjunction with layers and layer masks. We do find it useful to keep Smart Sharpen on default settings for everything except radius and amount, and treat it as a second Unsharp Mask tool, which preserves the sticky settings in both tools. This is useful because we generally use Unsharp Mask to achieve localized contrast, a sharpening plug-in or High-Pass sharpening for output sharpening, and Smart Sharpen for quick sharpening of screen images.
As mentioned earlier, we often prefer to sharpen on a layer with a High Pass filter using a blending mode such as Overlay, Soft Light, Hard Light, or Linear Light. This method concentrates image sharpening on the edges of shapes in the picture.
|Figure 6 Screenshot of High Pass Filter dialog.|
As you can see in this screen shot, the High Pass Filter has detected all the edges in the photo. By varying the Radius setting and choosing the Blend Mode, you have many adjustment options. Not only that, but since this is done on a layer, you can adjust the layer opacity as well. This sharpening technique was preferred by Bruce Fraser for output sharpening as he felt that it came closer to matching traditional drum-scanning sharpening than could be achieved with Unsharp Mask or Smart Sharpen.
Another edge sharpening technique can be accomplished by creating a duplicate layer, sharpening with Unsharp Mask set to a radius of 4-8 pixels and an amount of 50. This will oversharpen the image, but then you will set the blend mode of the layer to Lighten. This will have the effect of lightening all the edges, which has a similar effect to the High Pass technique. This technique can be used by itself, or as a second pass technique for medium or high frequency images that need some extra sharpening punch.
An advantage of sharpening plug-ins is that they have pre-defined parameters built in. This takes much of the guesswork out of output sharpening. You can choose an output media, and the software usually already has factored in the size. Push OK, and you have a ready-made sharpening solution. The PhotoKit Output Sharpener shown below saves the result in layers, so you can make additional adjustments if desired. It is good practice to create a duplicate layer before using other sharpening plug-ins, which do not automatically create layers. This gives you more adjustment options and allows you to save a "sharpening layer" for delivery files. Be aware that not all output sharpening plug-ins work on CMYK files. If they don't, you'll need to do this step in RGB mode and then convert to CMYK.
|Figure 7 PhotoKit Output Sharpener|
In general, photographers should be responsible for capture sharpening. In a batch output workflow, this will make the images appear reasonably sharp without creating artifacts that will interfere with additional image processing. Creative or process sharpening should be done by the post-production person, who may be the photographer, the retoucher, a graphic designer, or other person who is tasked with finalizing the image file. The biggest issue with regard to sharpening is output sharpening. Since output sharpening should ideally be done at the final reproduction size, this often means that it needs to be handled by someone other than the photographer. Unfortunately, there is not much awareness of how to sharpen images for output among graphic designers, clients, web designers, or even commercial printers. There are two basic approaches to dealing with this at present. The first is to take a defensive posture, and use an edge sharpening technique for images that you are reasonably certain will be printed same size or scaled down. Edge sharpening seems to hold its effect somewhat better than pixel sharpening does when images are resized. The second approach is to deliver images that have not been sharpened for output and to communicate the need for final output sharpening with a readme file or delivery memo. If you are delivering TIFF files, it is possible to create a sharpening layer and deliver layered files. The image receiver can choose to use the layer or not, depending on whether the sharpening effect holds up or not. Since edge sharpening techniques involve making a duplicate layer, it makes this option an easy one from a workflow standpoint.
Sharpen in InDesign?
One solution that we have advocated with Adobe is to have a sharpening function incorporated into the InDesign application. Since designers usually use InDesign to crop and scale image files, it would make sense to have a sharpening function as part of that workflow. Currently, to do sharpening properly, designers are faced with cropping and scaling for position only (FPO) images and then repeating that work in Photoshop, sharpening the images appropriately, and then re-linking the sharpened images to the InDesign document. Very few designers have the time or inclination to go to this extra effort.