VueScan and multi sampling

Discussion in 'Digital Darkroom' started by bernard_frank, Jan 1, 2007.

  1. For the VueScan users out there who scan b&w, is there any advantage/inconvenient in using multiple
    passes (what VueScan calls "samples") beyond, say, 4 or 6? Time notwithstanding of course. What do
    you use? My scanner is Nikon LS 4000 for 35 mm, and Epson 3200 for MF (which is on the
    verge of being replaced by something more state-of-the-art) on a Mac G5.

    Many thanks.
  2. Multipass scanning is used to reduce noise in particularly dense areas of negatives or slides through averaging. This will sometimes reduce the background noise level enough to see some detail otherwise buried in clutter.

    As implemented on the Nikon scanners, multipass scanning is done on the same line before indexing the scan head or film carrier (LS-8000/9000). It is accompanied by a slight but noticeable loss of resolution. Flatbed scanners usually "multiscan" by repeating the full-page scan, with a significant loss of sharpness (poor registration). However many flatbed scanners are so noisy that multi-pass scanning improves nearly any image.

    You should not need it for well-exposed B&W negatives, and never for color negatives (which have a relatively low maximum density).
  3. I use this Vuescan multisampling together with the long exposure pass on a Canon 9950F, scanning B&W negatives and color slides in 6x7 and 5x7 and by testing I found a big improvment over a standard scan.<br>
    I just looked for my test images but it seems I have deleted them so I cannot show, but for me multisampling with long exposure brings noticable less noise in dark areas and a little more in details there.<br>
    I didnt find a difference between 8x and 16x multisampling so i just use 8x, but this allways, IMO its worth the time.<br>
  4. I see no point in using more than 4 scans for averaging noise, if at all, even with a flatbed. Beyond that, there are diminishing returns not justified by the time needed and the bad effects of mis-registration. While I have experimented with multi-sampling on my Nikon scanners (LS-4000 and LS-8000), I have not found any salutory benefits.

    The degree of noise reduction is proportional to the square root of the number of samples. This is based on a reasonable, statistical approximation which is largely distribution-free. In other words, 4 samples will cut the noise in half. To cut the noise further in half, you need 16 samples. As a matter of experience, you need a factor of two in any parameter before you see a significant difference.

    While multi-sampling might help with DMax issues in an high-contrast film like Velvia, it can't recover "shadow detail" that simply isn't there. Shadow detail in Velvia, I'm convinced, exists only in the minds of its devotées. Multi-sampling is never necessary with well-exposed negative film based on DMax, but may help with over or under exposed images. Put your time and money into a good light meter or practice with the one you have, and learn the quirks of the film(s) you use.

Share This Page