How to merge a center focused scan with an edge focused scan?

Discussion in 'Digital Darkroom' started by russell_brooks, Jan 16, 2007.

  1. Problem 1:
    My Minolta 5400(1st version) cannot keep both the center of the negative in focus as well as the edges. This is because that even flat film has a slight bow towards the edges lying in the negative carrier.
    Idea:
    Make two scans. One focused in the center, one on the edges, then merge the two in Photoshop.
    Problem 2:
    I dont know how to do the merge. Can anyone help with ideas? I'm posting some sample images below so you can better see what I'm talking about. Also in my photo.net workspace Scanner Focus Problems you can download the original jpgs if you want to give it a shot.
    Thanks...
    00JYJH-34463784.jpg
     
  2. The negative is TMY. It was scanned at 5400 and then resampled bicubic down to 1/3rd of its size...
    00JYJK-34463884.jpg
     
  3. The technique is a version of masking and compositing.

    Import each image to a separate layer in Photoshop. Adjust the transparency of the top layer and nudge it to achieve perfect registration. The "Photomerge" tool may help. You may need to enlarge the canvas size to gain some wriggle-room. If you only change the focus target, the registration may not change between scans.

    Once registered. Make the top layer opaque. Create a mask (probably circular or eliptical, feather it, invert it if necessary and delete (erase) the unmasked portions.
     
  4. The problem is registration *does change* since the distance from film to lens changes with the focus. In this example it's not as extreme, but I've some frames that are more bowed and registration will become a real problem...
     
  5. Try CombineZ or Helicon Focus software. It's really designed for creating high depth of field from a series of (macro) images with slightly different focus. However, it automates all the things that are painful in Photoshop, like getting the registration exact, any change in magnification due to different focus, and it automatically masks for the sharpest available image source to make the composite final image.

    http://www.outbackphoto.com/workflow/wf_72/essay.html
     
  6. If you have tried Photoshop CS3 beta, then there is a Layer autoalign function thet might
    help.
     
  7. Put the sleeved film under a copy of "CS2 for Dummies" for 24 hours to flatten the film, worth ever penny from Amazon.com.

    From the 5400 preferences menue, select manual focus before scan. Pick a point 1/3 from center to edge or the most important point in the frame for focus. DO NOT allow the film to sit in the scanner and heat up .

    Scan the frame once to get the shadows right and another to get the highlites correct. Same frame, different exposure settings. Do one right after the other without moving the film. Use the exposure control before final scan.

    Now you have two scans, one with correct highlites, one with correct shadows. Open both in PS at the same time. Select the frame with the correct highlits. Ctrl-Shft- ~ will make an automatic selection of the brighter than 50% pixels or do the selection of your choice.

    Move the selection with the move tool to the darker frame while having the "snap to" function selected. It will drop in place perfectly.

    To get both frames to show, go to windows-view-tile.

    You now have HDR using one frame of scanned film. I will attach a sample. The windows were from an over exposed scan and the background from a normal.

    DO NOT move the film- Do one right after another and save as a separate file.
     
  8. Put the sleeved film under a copy of "CS2 for Dummies" for 24 hours to flatten the film, worth ever penny [...]
    LOL! True! I love analog solutions. You made my day.
    Rather than presuming the solution exists on the monitor, do that... or use a glass carriers (with anti-Newton glass one one).
     
  9. No glass carriers exist for the 5400 or I would have one. I can`t figure a way to make one either.
     
  10. Ronald,

    Use a http://www.gepe.com/english/113-txt.htm anti-Newton-ring glass slide mount.

    Although cutting the negative strips into individual frames is a nightmare, it's well worth the trouble for curved film. Put the Gepe frame in the slide holder of your 5400. It will focus perfectly or you can always resource to the manual focus feature.

    I have just a few of them but they have served me very well both in scanning as in the darkroom, although my LPL-4500 enlarger has a glass sandwich carrier for 4x5 that I frequently end up using for 35 and 120.
     
  11. Ronald, my negatives are flat. They've been under pressure for years. I've tried that method. The real problem in on the far edges of the 1st and last negative in a strip(cut into 5 or 6 negs). The last little up-lip causes the worst focus errors. Frames 2, 3, and 4 tend to focus fine. Try it once yourself and you'll see what I'm talking about. And as far as HDR, that's nice, but I dont really have that problem since my negs are well exposed and the Minolta has a high density range. The problem isnt a highlight/shadow issue but a focus one. I'll try and post a few more images later today as an additional example. BTW, I didn't expect this to be easy and I kind of figured most people will misunderstand what the real problem is...<br>The Helicon Focus and CombineZ5 are something that I will have to take a look at... hmmm...
     
  12. I've also taken a Gepe full frame anti-newton glass slide and cut out the little bit of plastic one side so I could "mount" the 1st and last negatives in a strip. To be honest the scan wasnt any better than the normal and still had some slight out of focus issues. I've tried to tackle this for some time now and I'm convinced that some kind of a multi-scan and then software resampling will be the best approach...
     
  13. I was referred to http://www.graficaobscura.com/depth/index.html a while back. If someone could break it down into layman's terms, step-by-step in photoshop, or a photoshop action, it would be greatly appreciated.
     
  14. Tried glass mounts and found them to be a real pain to keep clean. With the Wess AHX500K mounts, I find my 5400 scans' edge to edge sharpness improved to a point that it no longer bothers me.

    Your approach to merge two scans is problematic. After re-focusing the second scan, it will be misaligned from the first one. I have never been able to align them in PS. This product supposedly can do the trick with sub-pixel alignment, but I have not tried it.

    http://www.atpm.com/10.10/optipix.shtml
     
  15. Re:

    "Make two scans. One focused in the center, one on the edges, then merge the two in Photoshop."

    There is *some* depth of focus, precious little, but enough to raise the question:

    Why not make one scan, focussed at a compromise point between the center and the edge?

    Well, to qualify that: at a compromise point between the lowest and highest point of the film emulsion, wherever that may be.

    I also went with the Wess AHX500K mounts with slides, I believe after reading about them from Robert. Slides are the worst for curving in my experience, and those mounts tend to reverse-dimple the center. Determinining the nearest and furthest points is a little more of a challenge then, it's not a simple dome anymore. Vuescan's manual focus can help you 'map' the film, giving numeric readings at various points.

    Another thing to be careful of: if you have your film holder over to the left extreme, with the emulsion facing left (nearest to the lens/sensor), and start a manual focus, slowly moving the holder to the right (away from the lens), you will likely get 2 peaks. The first is at the filming backing, the second is at the emulsion.

    In my experience with that scanner, using the Wess AHX500K mounts, compromise focussing (the sweet spot takes some research and trial and error to determine), doing the focus manually frame-by-frame (I prefer to use the actual knob, ymmv), coupled with a bit of sharpening, you can get decent sharpness, nearly out to the corners, without loosing noticeable center sharpness.

    BTW, I'd really like to hear an interpretation of the link I posted above. I don't think it would that easy: the proportions of a scan seem to 'flex' when the focus is adjusted.
     
  16. FWIW, I tried the Helicon software yesterday. It works well with smaller files. But with 2 larger files(30meg+) it bogged down. I used my laptop with 1 gigabyte a 1.8 processor and virtual memory that should scale up to 3 gigs. Also it is not made for stitching together two scanned images so you really have to play around with it a bit. On the other hand I think it has potential. I'm sure they could make a template that works to make it easier to work with the constraints needed to stitch scans... The other program couldnt even open my Photoshop-saved TIFFs so I couldnt test it yet...
     
  17. Mendel, I don't understand what you meant by positioning the carrier to the left (of what?) and the need for manual focusing on the 5400. With the Wess mount, I find autofocusing at the critical part of an image works fine in almost every case.

    The technique described in your referenced article sounds interesting, but I have not tried it. But the biggest obstacle for merging two images in PS starts with whether they are aligned *perfectly*. Perhaps that can be achieved if the images are captured on a digital camera mounted on a tripod and the subject remains stationery. But it is not easy to create two perfectly aligned scans. From my 5400 experience, I can do that if I only change the exposure but NOT refocus. That works well for dynamic range extension by merging, but not for improving edge to edge sharpness. As soon as a scan is refocused, it will be out of alignment.

    I would be weary of sw that merge images "automatically", or without an explicit step to align the images first *before* merging. The sw is somehow averaging the misalignments, which can result in softer edges.

    All the tutorials on merging multiple images seem to dodge this fundamental problem.
     
  18. jtk

    jtk

    Nikon has that "end of strip" issue as well. Their strip film carrier solves the problem, though it's a flimsy device, fussy, and rare.

    I suspect the Minolta carrier could be modified for flatter film with narrow strips of plastic or teflon tape.
     
  19. Robert, I didn't explain very well. If you're facing the front of the scanner, it's lens is to your left, the light source to your right, the holder in the middle. As you focus the holder is being shifted to left or right. At least, this is the case, and is very noticeable, with the manual focus knob. As you turn the knob counterclockwise the holder moves left, and vice versa.

    I start my focus procedure with the emulsion mounted per instruction (facing left towards lens) and the manual focus knob cranked fully counterclockwise, which positions the holder at the left extreme of it's travel (nearest the scanner's lens). Turning the knob clockwise moves the holder to the right, away from the scanner's lens. The bar indicators lengthen to max when the scanner gets focus. Continuing (very slowly) to rotate the knob clockwise, so that the holder moves more to the right (further from the lens), the bars fall back a bit, and then lengthen to max a second time.

    From test scans, I get sharper results at that second peak. I speculate the first peak is at the plane of the film's back, and the second peak at the emulsion.

    As far as manual focus vs auto, I found Minolta Scan Utility autofocus inaccurate. Also, it will only focus dead center, and I *know* that is not the median point of my Wess AHX500K mounted slides. Wherever possible, I try to manually focus on a point roughly 2/3 the distance along a diagonal line from image center to lower left corner. I settled on that after mapping the topo of a few scans with Vuescan's manual focus, and just by trial and error. Also, I aim for an area with light uniform texture, which seems to offer the best elonganation of the bar graph, and likely the easiest focus for the scanner. If a light uniform texture is not available at that zone, I'll switch to lower right corner of the image.

    Regarding merging scans, I agree it would likely be problematic, or impossible. You can see the proportions/geometry of scans shift with different focus points. Merging them would be akin to making panoramics, or worse.
     

Share This Page

1111