UPDIG Guidelines, Version 3.0

Sharpening

Why sharpen? To get the best quality out of any digital image, whether scanned from film or shot with a digital camera, you have to compensate for the mathematically unavoidable loss of perceived sharpness that results from digitization. Likewise, whenever an image is reproduced in a medium that relies on dots (whether they are the halftone dots of offset printing or the pixels of a computer monitor), there is additional, process-specific softening that must be anticipated and neutralized. Fortunately, there are good algorithms that perform the needed sharpening. For historical reasons (W), the sharpening process is often called “unsharp masking.”

Capture sharpening: Capture sharpening should compensate for softness introduced during capture. The goal here is to have the sharpest, clearest image for a client to view or use. If you are shooting JPEG and are trying to minimize post-production time, it may work for you to find an in-camera sharpening setting that you like. If you put quality ahead of speed — and especially if you may want large prints later — it is best to do capture sharpening in post-production. If you are shooting RAW Files, you may find that using the sharpening controls available in your RAW processing program will work well for you. Sharpening with Adobe Camera RAW, for example, applies sharpening just to the luminance channel, which tends to minimize artifacts and noise issues. If you use another RAW processor, you should test the effect of sharpening on a noisy file to determine if the RAW sharpening adds to the problem.

Noise: All digital capture involves some level of noise. There are three situations where it’s essential that noise removal be a part of the workflow:

• When shooting at high ISO.

• When shooting images where the highest possible quality is required, such as for stock or in images destined for large prints (even at lowest ISOs).

• long exposures with early-generation digital cameras that do not have a built in noise removal feature.

If there are noise issues with your files that are more successfully corrected with a dedicated noise-removal program, it is best to sharpen after noise removal, and not before (that is, not in the RAW processor). One exception might be the new Bibble software which incorporates Noise Ninja into its RAW processor.

Process sharpening: Process sharpening is done at the end of whatever processing (editing, retouching, resizing etc.) is done to an image. Process sharpening corrects for the fact that changes in size, color, tone and color space all affect sharpness. It also may involve the photographer’s creative decisions about how sharp the image should appear, including sharpening selected areas of an image for effect. This step may be left to the client if the photographer is delivering a quantity of images that are not being individually edited.

Output sharpening: Output sharpening compensates for the characteristics of the final output device, and it is based on variables such as final size, intended viewing distance and reproduction technology (offset, inkjet, dye-sub, etc.) It is applied only as the final step before output. Once again, you may choose to use Photoshop controls such as Unsharp Mask, Smart Sharpen, or one of the many third party plug-ins. If you are creating a master file, you should save an unsharpened version, or save the sharpening on a layer, so that new sharpening can be done when the file is repurposed or resized. You never want to have to guess if a master file has been previously sharpened for output.

Practical considerations

It’s important not to over-sharpen at any stage. The image should always be able to tolerate additional sharpening. Mild over-sharpening imparts a surreal impression, while gross over-sharpening creates obvious halos around objects.

The Shore Temple of the Seven Pagodas; photo © 1985 by David Riecks, www.riecks.com.

In the example above, the first image is correctly sharpened for its size, the intended display medium (your screen) and the subject matter. The second image is over-sharpened, haloing the roof line with pale blue.

Who does it? In principle, all sharpening could be deferred until final output. In practice, there are drawbacks: If you leave all sharpening decisions until the output stage, some photos may be rejected for apparent softness. In other cases, poor prepress work could mean the images are reproduced soft. If an image is shot sharp, it should be shown and delivered sharp. Using three stages of sharpening is ideal, but not always practical for workflow. Photographers should deliver images that at least have capture sharpening applied. If the photographer is editing or retouching images, he or she should then also apply process sharpening. Photographers should only deliver images with output sharpening if they are doing the final pre-press work. If the client or printer is going to resize or repurpose the images, then the client or printer should be responsible for output sharpening. This should be clearly communicated when the files are delivered, perhaps with a ReadMe file. Some photographers like to deliver a final image file with output sharpening on a layer. By toggling the layer on or off, the client or printer can either use or redo the output sharpening — depending on whether the files are resized, retouched, or prepared for a different type of printer.

Which tool? Photoshop CS2 and CS3 have a Smart Sharpen filter which offers more control than Unsharp Mask because it can reduce sharpening amounts in highlight and shadow areas. There are also third-party plug-ins, such as Nik Sharpener Pro, Pixel Genius Photokit Sharpener, Focal Blade and others, that can fine-tune and automate the process. To accurately judge sharpening, the image should be viewed at 50% or 100% on screen. At other sizes, such as 33%, 66% or 200%, you do not see a true representation of the image; Photoshop is interpolating the pixels to create these views. It’s also important to remember that the effects of sharpening look different on screen than they do when ink hits paper, so print tests should be made, ideally on the same device type that will be used for final printing.