1. Reminder: Please use our affiliate links for holiday shopping!

Image Sharpening

Discussion in 'Image Processing' started by Brian Beezley, Dec 15, 2013.

  1. Brian Beezley

    Brian Beezley Mu-43 All-Pro

    Because I'm an undisciplined beginner, my images invariably need a fair amount of sharpening. So I've been investigating various sharpening methods. In addition to the usual Unsharp Mask, I've looked at RL Deconvolution in RawTherapee and several advanced methods in Image Analyzer. I haven't yet compared all of the methods. I find evaluation of image sharpness quite subjective. It's not always easy to decide which method or parameter values seem best.

    To compare two images, I switch between two aligned windows with a mouse click. In Chrome I create the windows by right-clicking on each image and selecting Open image in new tab. Offline I use two instances of IrfanView. I find the switched-aligned comparison method much more revealing than scrolling between two images. Any differences pop right out.

    [​IMG]

    I took this shot with a manual focus lens, the Zuiko OM 35mm f/2.8. I used the lens wide open to minimize motion blur, but this left me vulnerable to misfocussing due to beginner's neglect aggravated by the narrow depth of field. I tried several methods to sharpen this image. To process it, I opened the RAW image with Olympus Viewer 3 and set Sharpness to -2, the minimum setting. This still applies a little Unsharp Mask sharpening, but it is a very small amount. Then I generated an uncompressed, 16-bit, 1024x768-pixel TIFF image for further processing by various programs, including OV3 itself. For consistency I generated all JPEGs with Image Analyzer. I set JPEG quality to 100.

    [​IMG]

    I generated this image with Unsharp Mask in OV3. I used strength 250, radius 1.0, and threshold 0. I have not used this sharpening method much and I'm not sure I chose the best parameter values.

    [​IMG]

    For this image I used Restoration by Deconvolution in Image Analyzer. I used a Gaussian blur, but I forget what values I used for spread and number of iterations.

    [​IMG]

    Here I used RL Deconvolution in RawTherapee with the default parameter values.

    [​IMG]

    Finally, this image used Limit sharpen IIR in Image Analyzer. If you compare switched images, you can see that in this image the gold glitter is less sparkly (more realistic, I believe), object edges are cleaner, and the girl's freckles are more visible. The makeup artist's hair also has fewer bright strands, which I believe is more accurate. Currently this and First order IIR, a related method, are my preferred sharpening methods.

    I'd appreciate any comments or advice on image sharpening. Eventually I'll learn how to focus carefully and hold my camera steady. Until then, I need all the help I can get.

    Brian
     
    • Like Like x 1
  2. robbie36

    robbie36 Mu-43 All-Pro

    Sep 25, 2010
    Bangkok
    rob collins
    Out of your 'sharpening' choices, I have only ever used Photoshop. So a few comments I would make.

    1) In Photoshop, you should really use 'smart sharpen' as opposed to 'unsharp mask'. Smart sharpen can be set to 'lens blur' as opposed to 'gaussian blur', it offers more control with highlight/shadow protection compared to threshold as well as a 'noise reduction' slider in Photoshop CC.

    2) The way that the sharpening works is that 'radius' determines the pixel width of the sharpening halo and amount, rather surprisingly stands for the amount of sharpening. You need a wider pixel radius for sharpening in print that has a higher dpi than a screen.

    3) When sharpening in Photoshop it is best to do it on a separate layer and set the layer to luminosity. That is because PS sharpens each individual color channel separately which tends to result in color noise appearing the photo.
     
    • Like Like x 1
  3. Just Jim

    Just Jim Mu-43 Top Veteran

    941
    Oct 20, 2011
    Best to learn as many different ways to sharpen as possible. They just give different results. I use PS, so my advice isn't likely to be helpful.

    I'll use Blend mode Soft light or vivid with a high pass, and/or smart sharpen, unsharp mask, or Nik. I'll use them differently across an image to select out what I want, unless the threshold tool is good, and even then I may mask out areas. I rarely use one as a global sharpen in an image I want to spend time on. Decent shots that deserve some time, but not a lot, may get a global pass by with a couple targeted sharpening (eyes)

    Don't get too crunchy. imo, images these days are way too crunchy in general. No doubt we'll all laugh at it in 15 years like the old overdone soft focus when it just wasn't needed of the 90's late 80's.
     
  4. Dramaturg

    Dramaturg Mu-43 Top Veteran

    614
    Jun 7, 2013
    Ukraine
    Yevgen
    I like your Restoration by Deconvolution results, at least with this particular picture. In order to achieve similar results in PS you'd need to create several layers and do various tricks. I tried this filter in Image Analyzer but it doesn't seem to work for me. It would be nice if others also shared how they shapren their images.
     
  5. photo_owl

    photo_owl Mu-43 Regular

    164
    Nov 8, 2013
    so so much depends on the subject and the final image medium - print, projection and screen all carry their own needs.

    as already highlighted, there are a huge number of approaches to 'sharpening', and even more tools to implement these nowadays! My general preference is to sharpen using channels/layers in PS as Robbie references above - but I would have to check the differences between setting the layer properties at luminesence and selecting it through channels (it may, or may not, be the same thing!). There's also the use of the high pass filter on a layer combined with hard light mode as a quick, but remarkably effective and controlled approach to some image types (particularily landscapes).

    all this before you start bringing in the issue of sharpening what's not 'sharp' in the first place - whether by design, accident or equipement. Whilst Brian's upfront about these examples, the application of appropriate rescue sharpening techniques won't necessarily translate across.
     
  6. Brian Beezley

    Brian Beezley Mu-43 All-Pro

    There is very little documentation for Image Analyzer so you have to experiment. Here's how I use the deconvolution sharpener. First I select Gauss blur. Then I click Guess and let the program estimate the spread value. The result never seems to work well for me so generally I multiply it by about 0.7 and manually enter that value. Then I repeatedly press Do iteration while observing the image. Results seem best when the image still looks good after 10-50 iterations. In other words, if the image starts to degrade due to sharpening artifacts after only a few iterations, I lower the spread value and try again. Whenever you change the spread value, the image resets to the initial image (like clicking Reset). Once I come close to a spread value I like, I increase the 1 in the box next to Do iteration to something like 10 to increase the number of iterations per click. This makes everything easier and faster. I don't know if this is the best way to use this sharpener, but it seems to work for me.

    Brian
     
    • Like Like x 1
  7. meyerweb

    meyerweb Mu-43 Hall of Famer

    Sep 5, 2011
    For the moment. this is the only part of your post I'm going to comment on. The idea that the need for sharpening indicates some sort of error on the part of the photographer, or a problem with the image, is inaccurate. Sharpening is an intrinsic part of the digital workflow, and you'll find it employed by the best pros as well as the newest novices. The novices might not realize they're sharpening, but every in-camera jpeg engine applies sharpening as part of the jpeg conversion process. In general, the more novice the target market, the more sharpening is applied in-camera. Pro bodies perform much less sharpening, by default, on the assumption that advanced users are more likely to do it in post processing.

    Anyway, why is sharpening necessary? Thom Hogan says it better than I have the patience to: http://bythom.com/sharpening.htm

    His writeup is a little dated in terms of the resolution figures of cameras, but other than that is still accurate, and one of the clearer explanations I've seen. I'll add only one thing: Some of today's cameras have no anti-aliasing filter, or much weaker ones than is typical. Those camera's images will tend to require less sharpening than cameras with stronger AA filters, but will still need some to produce the best images.

    So don't beat yourself up because your images require sharpening. So do everyone else's. I think two of the most common mistakes I see novices making are not sharpening enough, or at all. And then when they learn what sharpening can do, over-sharpening images to near destruction.
     
  8. Brian Beezley

    Brian Beezley Mu-43 All-Pro

    I agree. It's just that my exposure setting and camera handling are still so primitive that my images need more sharpening than those of experienced shooters.

    I read Sharpening 101 quickly, but I think the reason virtually all images need sharpening is not due to quantization, as suggested, but something more fundamental. Whenever you change image size, you must generate image samples that weren't in the original image. In other words, you must interpolate. Interpolation methods with which I'm familiar generate a smooth curve, often one that passes through the original points. You pick interpolated values for the new samples off the curve. To avoid aliasing, the smooth curve must bandlimit the original image spectrum in a way that rolls off part of the desired spectrum. This causes the image to be soft. I view sharpening as an attempt to restore the original image spectrum (the part within the new bandwidth) after resampling.

    I save all sharpening until the image is at the final output resolution. You have to do it there anyway to undo the softness due to resampling. If you sharpened earlier, you'll be generating artifacts from the earlier sharpening artifacts, which doesn't seem helpful. This argument ignores any nonlinear processing between rounds of sharpening, which if used may well invalidate it.

    I added an image from the deconvolution sharpener in RawTherapee to my sharpening comparison.

    Brian
     
  9. tomO2013

    tomO2013 Mu-43 Top Veteran

    799
    Oct 28, 2013
    How do you find Nik Sharpener Pro in the Nik Collection ? I'm new to raw development and to be honest while I've paid up front for Photoshop CC, I find it incredibly confusing and don't really understand it very well.
    Nik looks to have taken the complexity out, although I'm sure you don't quite get the same result.
     
  10. TassieFig

    TassieFig Mu-43 Top Veteran

    534
    Oct 28, 2013
    Tasmania, Australia
    I do all my sharpening in Lightroom, it's very good and easy to do. And, it doesn't matter when you do it, first thing or last but imo makes more sense later. I tend to do it after cropping and basic adjustments but before any real fine tuning (if that is necessary). It obviously depend on the subject but generally I set sharpening to 50-60, radius 0.8, detail usually stay at default and mask at 85-95. I found masking makes a BIG difference to the sharpening success.
     
  11. Smashatom

    Smashatom Mu-43 Regular

    120
    Feb 24, 2013
    Stratford-Upon-Avon, UK
    Sam
    There's some really great video tutorials on YouTube on the official 'Nik Photography' account for sharpener pro and the presharpener, I now get a lot more out of these plugins since watching them, can't recommend it enough!
     
    • Like Like x 1
  12. hiimage

    hiimage New to Mu-43

    4
    Jul 8, 2013
    my suggest for beginners is that don't ever over sharpen image . sometimes it may have different result.
     
  13. meyerweb

    meyerweb Mu-43 Hall of Famer

    Sep 5, 2011
    That may indeed be a factor, but even images that aren't resized need sharpening. Thom's article is quite accurate, and you can find more scientific, more technical articles that will provide the math to prove it.
     
    • Like Like x 1
  14. mattia

    mattia Mu-43 Hall of Famer

    May 3, 2012
    The Netherlands
    For web use/general use, I find a touch of unsharp mask is usually enough (for people shots); I spend a little more time on things with a lot of detail, such as landscapes. The Nik suite is incredibly powerful when used as an integrated solution, and I strongly recommend watching a few of their tutorial vids on using them in order. The combination of the RAW presharpener with DFine noise reduction and Sharpener Pro - with very easy local adjustments and local sharpening, and guidelines on levels of sharpening to apply for a given print size and medium - is very powerful and fairly easy to use. I tend to 'oversharpen' images I'm processing for print, because the detail looks better that way on paper. On a screen it's too much.
     
    • Like Like x 1
  15. tomO2013

    tomO2013 Mu-43 Top Veteran

    799
    Oct 28, 2013
    Thanks for the update.

    I've finally decided on my choice of workflow and RAW processing. I've invested in Lightroom 5.3 + Photoshop, DXO Optics Pro and Nik collection.

    My workflow is as follows :
    * Generally I import into lightroom first.
    * Cull the images that I don't want etc...
    * Open the raw in DXO (which automatically does a form of adaptive deconvolution sharpening opening the RAW that is specific to your camera body , lens and focal length combination). So far I trust DXO's default 'lens softness' choices for most of the lenses that I own apart from the Panasonic 100-300.
    * I then do a few tweaks to the exposure, highlight recovery, shadow recovery and select camera body colour profile as Olympus, export a 16bit TIFF (generally these are the only options selected).
    * At this point in lightroom I'll open the exported TIFF in Nik sharpener pro and use the creative sharpening where and when I need it.
    * Before I go to print from now on I'll use the Nik sharpener Pro output sharpening to do a little output sharpening - as a result of reading this really helpful thread here :)

    I know that each image varies in the amount of sharpening required so a one size fits all is not realistic to give, but that disclaimer out of the way - those of you who also use/used DXO in a similar workflow, do you find that you still need to apply additional 'unsharp mask' in DXO despite the 'lens softness' setting before export. Or do you simply just trust all sharpening (presharpening, creative and output) to Nik / Smart Sharpen in PS. Looking at just the presharpening part of the workflow, so far I haven't managed to get the overall presharpening effect in Lightroom / Nik sharpener pro to be able to match the default sharpening in DXO . But that is most likely simply me and my lack of experience!
    I'm completely open to having my workflow torn apart (particularly in the sharpening area) if you guys have any suggestions that could help me improve it - or would care to mention your worklfow if you also use another raw processor in your workflow besides lightroom / ACR.

    Thanks,
    Tom.