Capture sharpening radius - D800E

Discussion in 'Digital Darkroom' started by kentshafer, Jul 19, 2012.

  1. Trying to apply what I learned from Jeff Schewe's book, when I use Lightroom to capture sharpen 12 MP pictures from a Nikon D700, I usually start with a radius of 0.5 for images with high frequency detail (landscapes or dog closeups for example), 1.0 for medium frequency, and 1.5 for low frequency (such as portraits).
    Is there any reason to increase those starting points when capture sharpening 36 MP images from a D800E since a given eyelash, pine cone, or whatever occupies more pixels? Should the radius be 3 times larger? 1.7 times (the increase in linear resolution)?
    I know there's no substitute for experience and good judgment, and there may not be any cookbook answers. Still, I would appreciate any information that might help improve my understanding of the theory on this point.
  2. I don't have a d800, so I can't speak from experience with that particular body, but when I occasionally process images from cameras with substantially lower rez than my d700, I find that the optimal capture sharpening radius doesn't change. I usually find that something around 0.7 or 0.8 pixel is close to optimal for all sensor sizes, at least at the capture stage. This fits with the fundamental purpose of capture sharpening, i.e., to counteract the blur that occurs at the individual pixel level due to the anti-aliasing filter, the demosaicing process, etc.
    However, in the next step, when doing creative sharpening, I find that the optimal radius does indeed vary with the sensor size, subject matter, etc. because of the exact effect you described.
    Obviously, the final output sharpening step depends on the output medium and content, and the settings I use are independent of the dimensions (in pixels) of the image.
    Tom M
  3. Kent,

    You might want to consider Bart van der Wolf's new (free) Slanted Edge tool and measure your optimum capture sharpening radius for your lens(es) and camera combination. You can find the discussion at After running a couple of my lenses through Bart's tool, I not only found the tool to be useful, but quite informative regarding the radius settings for capture sharpening in general. It removes the guess work, at least in the region of sharpest focus. The only downside is that it does take time to take the shots and put them all through the tool, but I believe the reward is worth the effort.
  4. Tom, thanks, that does make sense—that we would want to capture sharpen at the pixel level, so to speak, regardless of how many pixels a given sensor might have, so as to bring out the best that that sensor is capable of producing.
    And thanks, Bradord, for that link, though at first reading it's way over my head. But I will go back and study it some more. For starters, I guess I'll have to learn what "deconvolution" means.
  5. Kent,

    I agree that Bart's tool can appear a bit daunting at first. It is really not as hard as it first seems, and you don't really need to know about convolution integrals to use it. Here are the key steps to his process. First, download and install ImageJ and print a copy of Bart's target. ImageJ is very helpful with the point and area selections needed in Bart's steps 1 & 2. Load an unsharpened TIF file of the target image shot at the measurement focal length and f-stop. Select 2 points on the slanted edge with similar RGB values, get their coordinates (x,y) with the measurement tool and input these into Bart's Step 1. I usually use a magnification of 800x for this step on my screen. For step 2, I usually zoom in one more time and select a 1 pixel by ~100-200 pixel (not critical) area using the area select tool. Plot the points and list them in ImageJ and copy and paste them into Bart's tool, Step 2. Click on "Calculate sigma" in step 3 and you have your first data point which gives the desirable sharpening radius (sigma) for that lens setting. If you repeat this across a representative set of your lens settings, you will have complete characterization of your lens. I plotted my results (sigma vs. f-stop) in Excel.

    Good luck!
  6. Tom, I find that a much larger radius (in pixels, not in % of frame height) is needed on a higher resolution photo. Usually in my workflow, I'll do a very mild capture sharpening at low radius, and then do a final sharpening when all editing is done. Then I back off that last sharpening step, save to another file, resize to a low resolution, watermark, and sharpen to create the web version of the image. If I use the same sharpening radius on both the high and low resolution images, the low resolution image will appear grotesque. I have to use a much, much smaller radius.
    The optimal radius has to do with the width of the blur of a hard edge, which varies by the lens, sensor, accuracy of focus, and even shooting technique (e.g. tripod or not). I take a less quantitative approach to that that Brad has described, essentially blowing up the image on the screen enough to see the individual pixels very clearly -- extreme pixel peeping level, as it were. I then try to characterize the width of an edge in numbers of pixels radius. In my workflow, I take roughly 1/2 to 2/3 of what I subjectively call the edge width, and that is my starting value for sharpening radius. I may later tweak this value a bit, but that's where I start.
    Anyway, in direct answer of Kent's question, I would say the 36 MP images would need about the same sharpening radius as the 12 MP only if they are 1.73 (square root of 3) times as sharp (i.e. if the lens outresolves both sensors, all things considered, including shooting technique -- not likely). If both sensors outresolved the same lens, then I would say the 36 MP images would require 1.73 times the sharpening radius. The truth is likely somewhere inbetween, depending on the lens, etc.
  7. Hi Sarah - I think we are probably saying the same thing. I was just describing it in terms of the oft-quoted 3 stage sharpening workflow (*). Like you, my 1st stage (capture sharpening) is also always mild and of low radius. My 2nd stage (creative sharpening) depends on the image content, and, like you, my 3rd stage (output sharpening) depends on the output resolution, output physical medium, etc.
    Tom M
    (*) - "Real World Image Sharpening with Adobe Photoshop" (second edition) by Bruce Fraser & Jeff Schewe
  8. This fits with the fundamental purpose of capture sharpening, i.e., to counteract the blur that occurs at the individual pixel level due to the anti-aliasing filter, the demosaicing process, etc.​
    If my understanding of the OPs question is correct, the D800E does not have an anti-aliasing filter in front of the sensor. So, different rules might apply.

Share This Page