Unsharp mask and scanned negatives

Discussion in 'Digital Darkroom' started by john_simon|5, Dec 1, 2009.

  1. I am in the process of scanning my old negatives using a Coolscan 5000 (the Great Project). I use GEM in Nikon Scan to tame the grain in the first pass. I have been doing some minor tidying up on the scans using Photoshop - to which I am also new. In that process I have been using unsharp mask at the end before sending the happy snaps to iPhoto as JPEGs.
    At first I was using settings of Amount:125%, Radius: 2, with threshold varied to limit the extent to which grain in skies got emphasised. (I am scanning consumer 400ASA film - Kodak Gold or similar - that seems to be quite grainy.) While this seemed to work OK I was never really happy with the way it interacted with skies and the remaining grain. (Although this was an opportunity to learn about masks to mask off the sky when sharpening.)
    In my experimentation last night I stumbled upon settings of Amount: 50%, Radius: 4 with threshold 0, or maybe 1 and it seemed to be working like magic. Grain wasn't getting emphasised but features were getting sharpened nicely. The results seemed to be much less noisy but the sharpening was just as good.
    Can someone help me understand what is going on and why the large radius low amount settings might be working so much better for me? Is there a downside to these sort of settings that I have overlooked but will be terribly annoying as soon as someone points it out to me?
  2. That radius seems way too big for film scanned at 4000dpi. It's going to miss fine detail and I'd think you'd see some halos. I think I use a setting closer to .5 with threshold zero or 1, and often sharpened after reducing grain with Noise Ninja.
    These days I use Photokit Sharpener or Lightroom's sharpening tools instead.
  3. I agree, the radius does seem large, but it seems to work fine. So, to date, I haven't noticed it missing fine detail that has reasonable contrast - for example, even sunlit hair gets sharpened nicely. When I've tried low radii I just can't see it having much effect at all, and when it does it seems to emphasise noise as much as detail. At least at an amount of 50% I haven't seen any (objectionable) halos.
    Which is just a long way of saying - I know it shouldn't work in theory, but in practice it seems to. What's up with that?
  4. If you're viewing on an LCD, the zoom view will influence how you're sharpening appears. What looks good and crispy at 100% view may end up looking grainy at 25% view or when downsizing the file for web viewing. How sharpening appears on a print should be viewed at 25-50% zoom as well.
    You should post in this thread one of the sharpened negatives at 700 pixels on the long end using your 50 Amount/4 Radius to see what downsizing does and to get a second eye.
    I've often used the very same USM settings you've mentioned on certain images from my DSLR that contain fine detail to avoid grainy/dusty look when downsized from 3000x2000 pixels destined for the web.
  5. Here is a photo that I've treated as follows: Scan at 4000dpi on Coolscan 5000; ICE and GEM=4; Photoshop: USM at 50/4/0; Bicubic sharper scale to 700pix; Save for web at JPEG quality 90. Is that what you meant?
    When I was using 125/2 the pink sidings and roof would get very noisy at 100% view whereas with these settings I didn't observe quite as much of that. I can't say that I can see much at this scale but better eyes than mine might. I'll post the unsharpened file next for comparison (i.e. exactly the same steps as above but without the USM step).
    (Also hoping that I get the option to add the file at the 'success' page as this is my first time trying to post an image here.)
  6. Actually I had to set the quality to 63 to make it fit the 100KB constraint. Here is the original treated identically.
  7. I've been scanning some old trip negatives. Unfortunately a lot of the rolls of film were Kodak Gold 400 speed film. (GC 400-8) Probably what was available at most of the photo shops in India at the time. I'm using a Nikon LS-5000 as well. This type of film scans in horribly, and results in very noisy/grainy scans - unacceptable when compared to scans from other film types. The noise is worse in the lighter areas of the image (darker areas of the negative) - to the point where I think the scanner just can't handle this film very well.
    I suggest giving Neat Image or Noise Ninja a try. While they're not perfect, they each do a very good job at reducing the noise that you get when scanning this film. Try turning of all post processing except for exposure & Ice (i.e. turn off the sharpening that NikonScan wants to do) - then apply a noise removal tool (Neat Image or Noise Ninja).
  8. I think that whatever setting works for your photos is the one to use.
    I've read about the settings before, but forgotten what each setting means. I just leave the radius and threshold at whatever they're at (low) and vary the amount. If the amount required is too high, I divide the amount in half and do it in two passes to reduce noise.
    With 35mm unsharp mask really packs a punch because of the small negative size. Try a variety of settings on one photo and print, see how it looks.
    My rule of thumb, also, is to get it looking "real sharp" in the preview, then back it off a bit, until it's just "sharp enough", to minimize the effect on the grain.
  9. John,
    Both are excellent looking scans. I can barely tell the difference between the two but I much prefer the unsharpened version.
    However, you didn't mention previously you used Bicubic Sharper when downsizing so this may account for why the second looks more natural looking though the difference is subtle at this size in this thread.
    Also 4000 ppi is very high rez and so the radius can withstand a wider setting. You'ld have to zoom in at a level to see how thick one granule of film grain is compared to high contrast edge detail to really understand how far to go on radius.
    I'ld suggest you watch out using Bicubic Sharper because it can produce some odd artifacts when combined with regular sharpening and jpeg compression.
    Really nice color and texture in those shots.
  10. Oh, and just remember when sharpening landscapes like this, if you go too far on sharpening, it can make objects in the scene look like a miniature no matter how close or far away they appear in the scene. And that's what's almost happening on the first sharpened version. Might not want to use Bicubic Sharper and
    see if that helps.
  11. I think I over sharpened this image below not even using Bicubic Sharper.
    It really bugs me when this happens.
  12. Tim,
    Thanks for the comments. There is a bigger difference between the two shots before scaling down and the second one does have a bit of softness to it to my eye prior to scaling. I only chose Bicubic sharper because it was the option that was recommended in Photoshop for downsizing and I did't know any better.
    As for grain size - bigger than genetically modified puffed wheat. That sounds like a plausible reason for why large radii are working better for me with these scans.
    The colour and texture are what caught my eye as I was walking past that place - shame there was an electric railway gantry in the way (although that is the reason there is the cutting right under the house so you can't have one without the other).
  13. Christopher,
    I'm heartened to hear that my issues with this film are shared with others. It sounds like I'm in for a pleasant surprise when I get onto the more serious part of my archival project - scanning in my father's slides from 40 odd years of photography. Scanning my old happy snaps is just my warm up.
    I also think that it is not so much a case of the film scans horribly or the scanner has problems as the film just has issues - when I look at enlargements I've had done from this film the grain is quite evident. Viewing it on screen at 100% and 4000dpi just reveals things that were always there in the original but that were never noticed because one doesn't typically view an 11x14 enlargement from 2 inches.
  14. Hi John, I think you'll find the slides both easier to sharpen and also in need of more sharpening. Consider sharpening in 2 passes- the first on the 4000dpi scan and the second after you resize but before you print or upload.
  15. John, I have a method outlined below if you want to try it. It seems like a lot of work, but well worth it. This assumes you are working in 8 bit. If you are not, I'll let you know what to change in it. I'll also assume you are working from a flattened image. If not, you can stamp visible and we'll call that layer "Background".
    1. Figure out the change ratio of resizing the image. Divide the original long side by the output long side then multiply by 0.25. Example, if your original is 4000 pixels wide and your output will be 800 pixels wide, you would (4000/800)X0.25=1.25. Then blur your image at that radius using Gaussian Blur.
    2. Resize image with Bicubic.
    3. Make 2 copies of your "Background" layer. Name the layer above "Background" as "Sharpen" and the top layer "Detail".
    4. Set layer "Sharpen" to Luminosity blend mode and blur with GB at 0.3 radius.
    5. Select layer "Detail" and set mode to Linear Light. Then use Apply Image (under Image>Apply Image) with the following settings:
    Layer: Sharpen, Channel: RGB, Blend: Subtract, Scale: 2, Offset: 128 and leave all boxes unchecked.
    6. Smart Sharpen layer "Detail" at:
    Amount: 500%, Radius: 0.3, With More Accurate and Remove: Gaussian Blur.
    7. Smart Sharpen layer "Sharpen" at:
    Amount: 0%-100% (adjust to taste), Radius: 0.6, everything else the same as above.
    8. Merge layers "Sharpen" and "Detail".
    You can than add a layer mask to this layer to selectively sharpen where you need it.
    I have both an action and a script that will do this. The action doesn't do steps 1 or 2, but starts from step 3. The script will ask you the new long side dimension and do steps 1 and 2 for you also.
    As I said, it seems like a lot of work, but it's worth it in my opinion. Using the script is simple...all you will have to do is adjust the final Smart Sharpen dialog. If you process in 16 bit mode, let me know and I'll explain the difference in the steps.
    I hope I didn't forget anything. ;)
  16. Mike, could you post a before and after image demonstrating the significance of applying such a complicated procedure?
    It looks interesting.
  17. Sure. I grabbed a quick photo that I figured my have some decent detail in it. I may have to get something with a bit more detail, but you will be able to see a difference if you overlay one on top of another and switch back and forth or open each up in separate tabs and flip back and forth.
    Resized with no sharpening: http://www.nunuvyer.biz/Photoshop/Test/Out-Orig.jpg
    Resized with sharpening: http://www.nunuvyer.biz/Photoshop/Test/Out-Sharp.jpg
    I resized from 3888 to 1000 on the long side so I used an initial GB radius of 1.0 in the first step.
  18. Thanks, Mike.
    Yeah, that's not a very good sample image to see much of a difference over just applying two rounds of USM.
    I got the same results by applying a first round of 40 Amount, 2 Radius and second at 60 Amount, .6 Radius. Of course I worked off the downsized jpeg so those settings would be amped up for the original.
  19. Mike,
    Thanks for the suggestion. I will give your method a try but am unlikely to use it day-to-day. Some more information may make my circumstances clearer.
    I am scanning old happy snaps taken with a consumer P&S camera (Pentax 928) using consumer film (predominantly Kodak Gold 400). My destination for these shots is, predominantly, a JPEG file that will be viewed in iPhoto. (Some photos, however, were quite good and I intend to treat these more carefully with a view to printing some enlargements - these may benefit from your more involved process.) The destination resolution is undefined - at the moment it is onscreen at 1440x900 but it could be anything as the years go by and will undoubtedly increase. So I'm not downscaling these images, just sharpening to adjust for the inevitable analogue to digital softness one gets.
    My workflow is: Scan at 4000dpi 16-bit using ICE and GEM=4 and save as a TIFF. Open in Photoshop and tweak the photo - crop the margins, adjust levels and colour balance as required, retouch remaining dust and imperfections, sharpen and save as JPEG. This can be a pretty streamlined process unless I need to do more extensive work (I had fun working on a photo of me in shadow, in front of a glacier where no fill flash was used). The greatest problem in this process has been the fact that my film and scans are very grainy and if there is any sky in the photo the sharpening will accentuate the grain. I think this is accentuated by the way GEM deals with grain - you end up with a crystaline look to smooth areas - almost an oil painting. For the time being, using a larger radius seems to have helped. For the longer term, Noise Ninja may be on the Christmas shopping list. But in either case I'm looking for a near enough solution that is speedy. The rare keepers that I managed to capture will, however, get the white glove treatment.
  20. I haven't found GEM to be all that great. I find NeatImage to be much better at suppressing noise while retaining detail. In particular, the ability to build noise profiles for the particular film or digital sensor does contribute to noticeably better results.
    As for sharpening, almost anything will work better than unsharp masking. Take a look at FocusMagic.
  21. Tim,
    I haven't tried the USM on that same image to see the results, but I'd be willing to bet you are absolutely correct. This method seems to produce better results as the resizing ratio gets larger (final image a smaller percentage of initial). I've also found that it works better on higher contrast images. The image I chose was probably a poor choice for this, but I had a client on the way over so I didn't have time to dig through my stash. I'll see if I can post a better example later. I certainly don't wish to push this as a superior method, but merely another tool that can be used if the situation calls for it.
    I understand what you're doing better now. And what I outlined is more designed for a downsizing for web output sharpening. I think you're looking for more of a capture sharpen method. Specifically one that allows for removal of noise/grain. I've started messing with a method similar (though a lot more involved) to what I outlined above that gives control over an unlimited number of ranges of spatial frequencies kinda like an equalizer for an image. That, you'd have to use the script for. I don't have documentation for it but if anyone is interested, I can write something up quick and provide a link to the script file.
    But in the meantime, there is a program called Raw Therapee that might do what you're wanting. You can find it here:
    Good luck! :)

Share This Page