Sharpening - the ongoing Quest

Discussion in 'General Technical Discussion' started by Frits, Mar 31, 2005.

  1. I find that in post processing, sharpening is the most used and least understood concept.

    Since the years that I went digital, I have grown in my understanding of it, tested a LOT and I feel that I still can improve.
    I recently was snapped out of my blissful state of satisfaction by someone who showed my shortcomings (I had adapted the until then used method since about a year).

    My current workflow of my D1x images(with accent on sharpening):

    In camera:
    - shoot in RAW exclusively
    - no in-camera sharpening
    - WB: Cloudy
    - everything else: neutral

    In Nikon Capture:
    - adjust exposure
    - curves adjutment (largely following what I learned in RR's ebook)
    - if needed, punch up colors ever so slightly (few points only)with Color Booster
    - USM (66, 5, 0 )
    - save as 16 bit TIFF

    In PS CS:
    - in LAB mode, a USM pass in the Lightness channel (does wonders on the edges)
    - Back in RGB mode, apply a moderate pass with Fred Miranda's Intellisharpen II
    - Save final product as a 8 bit TIFF

    Previously, following my understanding of RR's ebook, I stopped after the NC stage.
    Now, adding the LAB lightness channel USM stage and a touch of FM's Intellisharpen does wonders. The projected, as well as the printed image has improved dramatically. Gorgeous, much sharper pictures (certainly not oversharpened).
    The only sacrifice is that the pristine noiseless D1x images (@ ISO 125) show a bit of induced noise, WHEN VIEWED AT 150% AND UP. I underscore this point, since I would view this as unacceptable in the past. I now realize that this does not show AT ALL when viewed on screen or in print. Of course, the pristine D1x files are so free of noise to begin with, that they can take this slight hit quite gracefully.

    So, this is where I stand today in my Quest. Surely, there are ways to further post process to address the slight induced noise, but I really don't care about it anymore, it is meaningless to me.

    Comments anyone? I'd love to hear what you think and what your experience is with the Quest for sharp, clean images.
     
  2. Flew

    Flew

    994
    Jan 25, 2005
    Alabama
    Fritz,

    I'm relatively new at this (been at it about a year), and I'm still very much in the learning mode, but my objectives are not unlike yours. I do most of my processing in NC, per my training from Ron, then do my cropping, sizing, final USM, and framing in PSP 9.

    Yesterday I got my copy of PS CS, so I start all over on the learning curve. I was not at all unhappy with PSP, but wanted to see what the Miranda plug-ins were like, so I've 'upgraded' to the high $$ spread.

    I'll let you know what I think in about a month, after I figure out where everything is. 8)

    BTW, I appreciate your post. It gives me one more data point in my quest for the perfect PP regime. :wink:

    Regards,

    Frank
     
  3. Frits,

    Great post with excellent information. Do you have any samplese before and after your "Quest" ?

    What do you mean by Lab Mode. I understand that there is Image/Mode/Lab color but what exactly is that and how beneficial is it to us ?
     
  4. Chris101

    Chris101

    Feb 2, 2005
    Arizona
    Hey Frank, congratulations.

    PSCS (at least for me) is as important to my photography as any of my lenses or cameras.You might want to check out Scott Kelby's book Photoshop CS for Digital Photographers. He's got a sound byte style, but also easy to follow cookbook instructions to do all sorts of things.

    You may opt to stay with NC for RAW conversion, but if you want to use ACR, read Bruce Fraser's Camera RAW with Adobe Photoshop CS. That book changed the way I expose as much as Ron's book changed how I hold a camera.

    Being used to Nikon Capture, you might want to open the Histogram from the Windows menu. Set it to Colors from the Channel menu, and to Expanded View and All Channels View from the palette menu (a little triangle in a circle in the top right of the histogram window.)

    But above everything else HAVE FUN!!! I can Photoshop for hours - it's worse than any computer game I've ever had.
     
  5. Chris101

    Chris101

    Feb 2, 2005
    Arizona
    Hi Frits!

    I have been working on perfecting a sharpening routine ever since I got my D100. Previously, my digicam made a presharpened JPEG file that didn't take additional sharpening very well. D100 files not only appreciate sharpening, they require it.

    I read about L channel sharpening in one of the references I put in my post to Flew above, and it worked well. But then in the other book I recommended, Adobe color guru Bruce Fraser said that converting from RGB to Lab and back, weakened the continuity of the colors in a 16 bit file, and restricted further editing.

    This may be the cause of the degradation you see at high magnification. I haven't stopped going to Lab-land (for more than just L sharpening, like monochromizing noise for example), but I do consider the effect on, especially shadow regions of, the photo.

    One work-around I do, especially for people pictures, is to (over?) sharpen the red channel, especially in the facial features and hands (or other body parts you wish to emphasize.)
     
  6. Once I discovered PixelGenius PK Sharpener, I haven't used anything else but. You paint it on where you want the sharpening, and you can adjust the opacity/level of sharpening. I just love it, and it marries with my IntuosII tablet very well. I think they've got a trial on their site. Highly recommend it.
     
  7. MontyDog

    MontyDog

    Jan 30, 2005
    #1064 - You have an error in your SQL syntax;
     
  8. MontyDog

    MontyDog

    Jan 30, 2005
    #1064 - You have an error in your SQL syntax;
     
  9. I'm happy with a touch of PSCS USM. Before that I add a touch of PSCS bicubic smoother which reduces noise artifacts and works well with the USM sharpening.
     
  10. I'm happy with a touch of PSCS USM. Before that I add a touch of PSCS bicubic smoother which reduces noise artifacts and works well with the USM sharpening.
     
  11. The LAB mode explanation was addressed before.
    Here is an example (NOTE: These are 100% crops, taken from an untouched RAW file). Also, internet viewing can take you only so far. The original, full size images are soooo good!

    My "old" way. Again, this has been perfectly satisfactory to me for about a year. Looked fine and printed beautifully (remember, this is a 100% crop):

    [​IMG]

    My "new" way. Dramatic improvement, totally spectacular on screen / in print.:

    [​IMG]
     
  12. Iliah

    Iliah

    Jan 29, 2005
    nowhere
    When using USM on RGB, it is possible to avoid colour artifacts using Luminosity blending or Fade to Luminosity.

    Round-trip RGB-Lab-RGB of course adds rounding errors :) , but they are more or less the same as noise levels. I can only speculate that in CS2 those errors will be significantly smaller, and Lab will be much more usable then now. I can also expect that to use CS2 to full power some additional hardware investments can be necessary.

    Generally, sharpening tends to work better in colour spaces with low gamma TRC. That partially explains why sharpening in Lab works so good.

    Sharpened layer blended on darken or lighten modes helps avoid halos.

    Using edge mask helps to avoid sharpening flat and OOF areas.

    Here is a small variation, based on suggestions made by Dan Margulis.

    Open an image, and duplicate it. Apply USM to the copy 400/2/4. Now Fade 100%/Luminosity. Activate original, and use Apply Image, Source to USM'ed copy, Target to original image, Blending should be Darken, Opacity - 100%. Now activate USN'ed image, and use Apply Image, Source set to original image, Target - USM'ed image, Blending mode - Lighten, Opacity - 50%. Without saving the original image, re-open it to compare the results :)

    Of course, the above can be done based on layers, and in Lab colour space, also using masks for edges and counter-masks for areas without textures and bokeh areas (BTW, blur works better in high-gamma spaces, so to prepare blurred masks sometimes it is needed to raise gamma - this way amount of blur needed to mask noise and make smooth transitions is smaller).
     
Loading...
Similar Threads Forum Date
Sharpening for web? Need advice General Technical Discussion May 30, 2012
Black art of sharpening General Technical Discussion Dec 15, 2009
Sharpening Images Post Processing - D90 General Technical Discussion May 10, 2009
Sharpening experiment General Technical Discussion Jul 20, 2008