# Does the order of sharpening and noise reduction matter?

I have been shooting for more than three years, but I never do too much in post-processing. I tend to be strict and keep in mind that PP can only do little to save an otherwise-great photo.

I was going to enter the "Photo of the week" thing, and I was doing sharpening and noise reduction in Photoshop.

NR makes your photo blurry, while sharpening amplifies noise.

I wonder, does the order of these operations matter at all?

Should I apply noise reduction and then unsharp mask, or the other way round?

• possible duplicate of Why should my last post-processing step be sharpening? Nov 14, 2011 at 13:22
• The suggested duplicate, though, doesn't ask about this specifically, and the existing answers don't go into detail. Nov 14, 2011 at 14:14
• I won't disagree with any answers, because I don't know the details of the Photoshop algorithms, but felt it's worth pointing out that the answer depends on the nature of those algorithms. If they are linear, and there's no clipping involved, then the order does not matter. The simplest sharpening and noise-reduction algorithms are indeed linear, but I suspect Photoshop (and many other packages) would use more sophisticated nonlinear versions. For nonlinear operations, there's no substitute for actually trying them in both orders and comparing :-). Nov 14, 2011 at 22:47
• Possible duplicate of photo.stackexchange.com/questions/2318/…
– mmr
Nov 18, 2011 at 5:20
• @whuber: Good point, however its probably worth while pointing out that most noise removal programs (including Photoshop/Lightroom) don't actually publicize the specifics of their algorithms. You could make an educated assumption, but it would be difficult to know for sure. Given that with linear algorithms order does NOT matter, its still a safe bet and best practice to deal with noise early. Nov 21, 2011 at 4:34

Simple answer is YES!

Noise is an undesirable artifact, and many operations can enhance its appearance, not just sharpening. Tweaking curves, adjusting contrast, working exposure, etc. can all have some impact on the noise that is present in an image...although sharpening tends to have the greatest impact. It is important to handle the bulk of your noise reduction as one of the first steps of your development workflow. You should not necessarily aim to "eliminate" all noise...rather your goal should be to reduce the effect noise has on an image as much as possible without adversely affecting detail that you wish to keep. The tool you use to remove noise can have a HUGE impact on the results as well...Photoshop in particular is not really known these days as having the best noise reduction tools. Lightroom has some fairly excellent noise reduction, and there are other tools like Noise Ninja and Neat Image that also do an excellent job.

Keep in mind your final presentation format and size...a LOT of noise that may be present in an image at 100% is going to get "absorbed" (for lack of a better term) into the final presentation format. Prints can sometimes exhibit pronounced noise, particularly when the noise contrasts well with the base image. Usually a moderate amount of noise, particularly in smooth gradients, can be beneficial to final print output as it normalizes content and eliminates harsh transitions. Presentation on a computer screen is often best done at significantly lower resolution than the original image, and downscaling is a superb mechanism for eliminating noise. If you start out with a moderate amount of noise, or an even distribution of high amounts of noise, downscaling for display on a computer screen (i.e. via a web page) can be all the noise reduction you need.

Sharpening is a facet that is closely linked to the viewing medium of an image. Generally speaking, its best to save sharpening for last. Not only that, its best to keep sharpened images saved as separate files from your primary work image (which in turn should be separate from your original master image.) The amount of sharpening done, and the type of algorithm used and its settings, should be matched to the output medium for your photos. If you intend an image to be viewed onscreen, the kind of sharpening you do will be different than if you intend to print. The size of the final image for either medium will also affect how much sharpening you do, and how that sharpening is done. It's a good best practice to save resizing, cropping and sharpening for the very last two steps of your image editing. Duplicate, crop, resize, sharpen. If sharpening greatly enhances the remaining noise in your image, you may also want to do a final "light" pass of noise reduction to minimize it as much as you can without adversely affecting sharpness.

• An important note about downscaling...its probably best NOT to rely on "quick and dirty" scaling previews that many programs offer (such as Lightroom). A lot of these quick previews use the most basic scaling algorithms, such as nearest neighbor or a highly optimized low precision form of bilinear filtering. Such algorithms are generally not very good at demonstrating the noise levels of an image that has been properly scaled down. If you wish to see how noisy your images might look when downscaled for presentation on a screen...actually do so, and see how things look. Nov 21, 2011 at 4:39

Always do sharpening last - if you do sharpening and then NR, you are going to first amplify noise and then undo the sharpening (because, as you already know NR will cause some blur)

Also, sharpen for your target medium (big jpeg, small jpeg and print all require different amount of sharpening) and make sure sharpening only happens once, for example if you upload to Flickr they will sharpen your images for you - the same is true for most photo sharing sites and printing labs, so unless you are using pro services that let you turn off sharpening or host/print the photos yourself you never need to sharpen.

• You should reconcile with your "e"... ;-)
– ysap
Nov 14, 2011 at 16:00
• @ysap - sorry, I typed this on my iPhone (I hate using this tiny on screen keyboard)
– Nir
Nov 14, 2011 at 20:48

Putting NR first is important, but there is a even more significant issue underlying your question. Doing little to no post-processing indicates that you shoot jpeg, not raw format images.

The post-processing software that pros use, Lightroom, Bibble, etc., does a far better job of producing high-quality images that any in-camera software.

• When you shoot a jpeg, the bit-depth is reduced from typically 12-14 bits per pixel to 8 bits per pixel. That results in loss of tonal detail, especially in the highlights and shadows. Post-proicesing a raw image gives you the ability to pull detail out of the shadows and restore it to the highlights before producing the jpeg.
• The colorspace is also baked into the jpeg, whereas you can change it in post with a raw image.
• Jpegs are compressed with a lossy algorithm, so additional image detail is lost that is present in the raw file.
• As post-processing software improves over time, you can reprocess the original raw file to get even better results.

Shooting jpegs in a digital camera is similar to shooting a Polaroid instant print camera. All the jpeg/print making decisions are made for you and the result is a heavily modified version of the image that could have been captured with raw digital or film. Yes, I know you can tweak some parameters in camera that control the look of the jpeg, but they are crude controls. Yes, you can refine these settings to work decently in a studio setting, but even there you can get far better images by shooting raw and post-processing.

If an image is worth capturing, it is worth capturing in the best format you can to preserve as much information as possible given the equipment you are using.

• I think you have made an assumption and based your answer on that. Sorry, I shoot in raw, and I already know all the things you listed, no offence, but this was not what I was asking about. Thank you for the input but your answer is a bit off topic. Peace. Nov 15, 2011 at 8:27