Following on from Tony's method here's mine - under a new topic as it has
nothing to do with Polaroid 120s. I don't think I'm too good at the actual
image processing side so I'm going to brush over a lot of that and describe
the overall workflow.
I am aiming to have scans of anything vaguely useful so first of all when I
receive my slides back from processing I do NOT open the box until I am
ready to scan the lot. I then batch scan everything (unless it is really
obviously bad) straight out of the box at 4000dpi 48bit portrait format with
Vuescan. This is before I really look at them - I know its hard - but its
the best way to beat the damn dust. Sadly it's too late for my older slides
and I just have to work round the dust/scratches and muck in other ways.
I tend to try and match the crop number from Vuescan to the slide number in
the box - it helps with cross referencing back to the originals. To do this
delete all the crop files in your Vuescan directory and if there is a slide
"0" scan it last and rename it "crop0000.tif". Where you have slides you do
not want to scan create a place marker file "crop00nn.tif" so that the
numbering keeps in line. Next delete the place markers and move the rest to
a new directory for this box of slides. If I'm starting with a box of
largely bad slides I just scan the OK ones and rename to get the right slide
number.
In order to minimise the dreadfully slow processing of 100MB TIFFs, I then
do a photoshop batch run through the whole lot reduce them to 600x400 8bit
and save them as a jpeg.
At this point I weed out and delete any rubbish.
Then I attempt to do overall colour balance (sometimes straighten horizons)
on the small jpegs (use record actions for later processing of the TIFF).
Sometimes if things are particularly problematic I will rescan the slide
with Vuescan or Silverfast and remake the jpeg. Once I'm happy with the jpeg
I despot the TIF and perform the processing I did to the jpeg to the TIF.
Crop as required but ensure I don't crop something out that I later might
decide I want. For absolute favourite images (hopefully at least one in
every box) I save a 16bit copy for archive because one day I may figure out
how to process it properly.
I then try do something to minimise the film grain/CCD noise. The most
effective way to do this is to down sample but I am trying to avoid this as
it is really a one-way process and CD-Rs are pretty cheap. If in doubt I
stick with 4000dpi, but for higher speed film I usually down sample.
The next bit I am not too sure about, but until someone comes up with
something better or points out some really nasty side effect I quite like
it.
The main way I minimise the noise is with gaussian blur - below radius 0.9
the effect on sharpness is not too drastic. My favourite technique at the
moment is to use a 3 layers. The bottom layer I sharpen quite aggressively
ensuring the unsharp mask threshold is set high enough to avoid sharpening
the grain/noise - even a little over sharpening is good be bold (see later).
I tend to sharpen it twice once about 70% radius 1.4 - 1.8 threshold about
12 and then re-shapen it 100% using threshold 6 . The middle layer I
gaussian blur about 0.8. I then blend the lower 2 layers at around 40-60% to
get the desired result (some over sharpening is disguised here). The top
layer I leave as the original to compare before and after results. When I am
happy with the result delete the top layer and merge the bottom 2 to get a
sharper less grainy image.
Now I have an image I am "happy" with I save a TIFF and overwrite the jpeg
with a new small jpeg from the TIFF. The jpegs still help with locating
images quickly - I think one for web use would be a good idea. Group the
images into sets for burning to CD-R add a text file (not word KISS) index.
I burn 2 CDs (on two different brands) one copy I keep and one I send/take
to Mum (250 miles away).
>From my CDs I can produce images at required sizes relatively quickly.
I keep meaning to add a simple html viewer to quickly scan the contents of a
CD or more likely a directory to help locate images. I also quite like the
idea of a CD case sized thumbnail printout - but I haven't got round to this
either. Any offers/suggestions for these will be gratefully accepted.
Steve
----- Original Message -----
From: "Tony Sleep" <TonySleep@halftone.co.uk>
To: <filmscanners@halftone.co.uk>
Sent: Thursday, June 14, 2001 7:07 PM
Subject: Re: filmscanners: Sprintscan 120 and new negative proile scheme
> On Wed, 13 Jun 2001 11:28:21 -0600 Michael Moore (miguelmas@qwest.net)
> wrote:
>
> > Tony: Would you be so kind as to give a step by step outline of your
> > technique
> > for dealing with color neg from exposure to final output? Am
> > particularly
> > interested in how you are dealing with 1. inversion...do you do it with
> > the
> > scan software or take it into PShop 2. setting white/black/gray
> > points/ etc.
> >
> > Thanx
> >
> > Mike M.
>
> I don't actually have a single regime, but a rather variable recipe which
> I adapt ad hoc depending on the problems that emerge. But I'll try and
> give an idea.
>
> First off, nowadays I invariably use Vuescan for colour neg, scanning to
> 16 bits. Having messed about plenty with Insight, Binuscan and Silverfast,
> I've found that whilst all of them can give very good results a lot of the
> time, each can occasionally result in a scan which an utter b*tch to sort
> out. Vuescan just seems more consistent, or at least I have evolved a way
> of working with it which works reliably for me. But this route is slow,
> far slower and requiring vastly more effort from me than the others. This
> suits me because I would rather scan once then do things incrementally.
> With the others, any significant problem I can't fix usually means
> re-scanning.
>
> I aim to do the gross colour correction in Vuescan ('cos it's rather good
> at it), but leave levels, saturation and final tweaking of curves and
> colour to be done in PS. VS handles mask removal, so I don't even need to
> think about that.
>
> Typically this will mean using VS with 'white balance' selected, but
> sometimes it isn't the best choice. This is just a trial and error thing,
> based on the preview from memory. Whatever is closest to ballpark is best.
>
> I'll select VS image controls so I get a scan which has headroom at both
> ends - ie from dark grey to pale grey rather than max.black to white
> highlights. VS default white point setting is too high for me, so I reset
> it to 0.01. I want to try and get everything off the film at this stage
> and make those decisions later in PS.
>
> Typically the VS output scan will look washed out, low contrast and
> desaturated as a result. This is good! With 16bits, there's plenty of room
> for improving things.
>
> First job in PS is to open the VS scan000n.tif file and do all the tedious
> spotting and damage correction then save the image over itself. Then again
> immediately to a different name/location. That way I can always go back,
> or create another version with different corrections for a layer. This is
> often the easiest way to get good highlights and good shadows in one image
> - two separately corrected scans from the same VS original.
>
> (Spotting is why I hate to have to go back and re-scan - it means
> re-spotting and that takes ages and is criminally boring).
>
> With a scan that is otherwise fairly correctly colour balanced, I'll then
> set the levels. I'll clip the black point slightly, leave a bit of
> highlight headroom, and get the overall gamma about right with the midtone
> slider. Sometimes all that needs doing after that is to increase
> saturation - I usually have to dial in +30 to +40 or so. Other times, I'll
> need to revisit levels (or contrast/brightness) as well - it just depends.
>
> Logically it would seem more sensible to increase the saturation as the
> very first step (to make colour errors more obvious), but I find I can
> never get it right if I do it before levels and have to adjust it again
> anyhow.
>
> With a scan which is 'off' regarding colour, there are various things I'll
> try depending on what I think will work best. Usually I'll start with
> levels again, and the channel histograms can be useful. I generally fix
> the black point first using the slider, again clipped a bit. What happens
> next is a bit suck it and see. You can mess about with the midtone and
> highlight sliders on each channel, but this can result in chaos. If that
> sort of thing is necessary, I find curves more intuitive and precise.
>
> A useful shortcut to correcting casts is to double-click on the PS
> highlight tool and set the tool to the tone and colour you want to
> achieve. For example if you have a bit of white shirt collar which is
> looking a murky pale blue/cyan, you'd select a neutral near-white. Drop
> that on the offending bit of collar and PS will adjust the whole image :
> magic! (though it can take some experimentation with the sample area, and
> the precise tone/colour you want). This works particularly well for colour
> negs shot in flourescent or tungsten, but it's best IMO to leave some
> trace of the illuminant colour - fully corrected just looks wrong.
>
> You can do the same thing with the shadow and midtone droppers, but I find
> the highlight one usually the most helpful.
>
> After getting the colour more or less balanced, I adjust the saturation
> and then make any final adjustments to levels, colour balance etc.
> That's it.
>
> Except it isn't (oh, I love the history list:). I fairly often run into
> trouble with levels and end up using curves instead of crude midtone gamma
> setting. Where there's a lot of important shadow detail, straightforward
> levels adjustments tend to extinguish too much. Then I will edit the curve
> shape to preserve the shadows. You have to be delicate here and realise
> than any separation you add to shadows is robbed from elsewhere in the
> tonal range - easy to get great shadows and flat, lifeless midtones and
> highlights.
>
> If none of that has worked satisfactorily, I'm going to end up doing one
> version which works in the shadows/midtones, and another with good
> highlights, knock both down to 8bits and then combine them as layers.
>
> Last stage is reduce to 8 bits, caption and save. Unlike most people I
> don't usually keep the high-bit scan at all. It has served its purpose,
> and I never want to see it again :) However I invariably do *all* the
> above at max optical resolution and keep the final version. Anything
> smaller gets derived from that.
>
> It'll be quite interesting to hear how this differs from other peoples'
> methods. I have no qualms about admitting I'm far from being clever with
> PS, the above has just fallen out of masses of trial and error, usually
> against the clock.
>
> Regards
>
> Tony Sleep
> - Online portfolio & exhibit; + film scanner
> info & comparisons
>
