Vuescan appears to be using "nearest neighbour" for downsampling

[simplicity]

Missed a portion of your question. Specific features I like:

scan-from-disk concept

What is this?

simple, fairly efficient, crop and rotate

Can certainly be done in PS.

scanner calibration with targets (just considering this)

What is this?

fairly good white balancing

Is VueScan controlling the scanner hw, or applying sw?

color neg film advanced workflow (just considering this)

What is this?

Note, I'm by NO means a pro. I'm sure there are other avenues for most
of the above, I'm just not that advanced/knowledgable, and Vuescan,
atleast some of it, works for me.

The key is that it works for you.

 

[Don]

Missed a portion of your question. Specific features I like:

scan-from-disk concept

simple, fairly efficient, crop and rotate

scanner calibration with targets (just considering this)

fairly good white balancing

color neg film advanced workflow (just considering this)

What struck me was that you do use Photoshop in postprocessing, and
all of the above (except calibration) are really editing tasks. I mean
even scan-from-disk is basically a fancy name for applying assorted
editing functions to a raw scan which can not only be done in PS but
PS has many more and better tools.

Note, I'm by NO means a pro. I'm sure there are other avenues for most
of the above, I'm just not that advanced/knowledgable, and Vuescan,
atleast some of it, works for me.

Yes, a familiar UI is certainly an important point (although I don't
think VS really has a UI) but if it works for you, that's what counts.


BTW, I also forgot to ask last time. Why do you downsample? Is it just
a question of file size?

The reason I ask is because I archive my stuff at optical resolution
of the scanner. If my scanner were capable of 5400 I would certainly
like to archive at that resolution even if that meant burning a few
extra DVDs. After all display resolutions keep going up...

Don.

 

[Mendel Leisk]

What is this?

Read Vuescan help file, available online.

Can certainly be done in PS.
Yes.


What is this?

Read Vuescan help file, available online.

Is VueScan controlling the scanner hw, or applying sw?
Software.


What is this?

Read Vuescan help file, available online.

The key is that it works for you.

Yup! Well, until I get a nasty little suprise, like the downsample method it uses.

 

[Mendel Leisk]

What struck me was that you do use Photoshop in postprocessing, and
all of the above (except calibration) are really editing tasks. I mean
even scan-from-disk is basically a fancy name for applying assorted
editing functions to a raw scan which can not only be done in PS but
PS has many more and better tools.


Yes, a familiar UI is certainly an important point (although I don't
think VS really has a UI) but if it works for you, that's what counts.


BTW, I also forgot to ask last time. Why do you downsample? Is it just
a question of file size?

The reason I ask is because I archive my stuff at optical resolution
of the scanner. If my scanner were capable of 5400 I would certainly
like to archive at that resolution even if that meant burning a few
extra DVDs. After all display resolutions keep going up...

Don.

Downsample was not an easy decision. I know full well, 'bout the time
I finish, hard drive and dvd size will be doubled.

I'm scanning somewhere north of 2000 slides. With downsample to
4000dpi (and crop of black edges), they'll JUST about fit on my
(mirrored) 250gig drives. That is 16 bit rgb tiffs. They are
grainy/handheld/mixed quality. I want to do them all, don't want to
cull. At 4000dpi, it will take about 50 dvd's for the raws.

I looked long and hard at the difference, and decided I could live
with the reduced zoomability. Usually, zoomed into 100% at 5400dpi,
I'm looking at some very well defined grain on a softening subject.

 

[Kennedy McEwen]

I'm scanning somewhere north of 2000 slides. With downsample to
4000dpi (and crop of black edges), they'll JUST about fit on my
(mirrored) 250gig drives. That is 16 bit rgb tiffs. They are
grainy/handheld/mixed quality.

Mendel, if they are grainy images then there is little, if anything,
being gained in saving these as 16-bit per channel. (Assuming that is
what you meant instead of 16-bit 64k colour (ie. 5-6-5 r-g-b) tiff,
since the numbers tie up at 4000ppi 16-bpc with each file around 125Mb).
You may as well half the file size by archiving them in 8-bit per
channel. You will also find that they compress much better as well,
using lossless LZW, when only 8-bit for the very reason that all of that
additional bit depth is just noise - not scanner noise, but noise on the
actual image recorded on film.

I would recommend scanning in 16-bits and then contrast stretching the
image so that the peaks are retained without clipping (so that no useful
image information is lost). Then reduce the bit depth to 8-bits and
save. The Minolta software probably has a function to implement the
stretching with minimum clipping in any case, making this a simple
operation. But if you use Vuescan then the same capability is there and
you can completely avoid the downsampling issue.

I suspect that using this approach you will get more images stored at
full resolution than you will achieve by downsampling and storing at
16-bits per channel, simply due to the improved efficiency of the LZW
algorithm once the noise content has been reduced. It all depends on
how grainy the images are to begin with though.

 

[Mendel Leisk]

I suppose the reason I'm saving 16 bit per channel red/green/blue is to
stick with my Vuescan scan-from-disk workflow. I appreciate that if I
convert to 8 bit I've instantly cut my file size in half.

Back a few years, I tried converting some of my Vuescan Raw Files from
16 bit to 8 bit and found degraded smoothness of tonality when doing
scan-from-disk.

The files I'm working with now are Minolta's 16bit linear output, which
(I believe) are akin to a Vuescan raw file, and can be used as such.
Accordingly, I shy away from converting them to 8 bit, or doing any
adjustment of histogram. I suspect scan-from-disk output from an 8 bit
conversion would show degration of smooth tone.

The earlier Vuescan raws I produced were of Tri-X, scanned with a Dual
II. With them, LZW usually compressed around 4/5. Now, with my Elite
5400 and 16bit linear files, LZW compression has a negative impact on
size, they go up in size. Quite a bit if I do the process thru PS,
marginally if through Vuescan. The latter by doing scan-from-disk and
outputting a new raw file with lzw on.

I gather the poor compression is due to the grittty nature of the
scans. With my earlier lzw compressed Vuescan raw files from Tri-X, I
could always tell when I'd done a gross mis-focus. The lzw compression
ratio shot up.

 

[Kennedy McEwen]

I suppose the reason I'm saving 16 bit per channel red/green/blue is to
stick with my Vuescan scan-from-disk workflow.

Why do you consider that to be important?

Scan from disk is no more than an alternative input into Vuescan - in
fact it is just another name for the File Open function in any image
editing application. Vuescan is a data capture application, and
performs those editing functions that it provides with less capability
than a proper image editing application - as you have now discovered.

I appreciate that if I
convert to 8 bit I've instantly cut my file size in half.

Back a few years, I tried converting some of my Vuescan Raw Files from
16 bit to 8 bit and found degraded smoothness of tonality when doing
scan-from-disk.

The files I'm working with now are Minolta's 16bit linear output, which
(I believe) are akin to a Vuescan raw file, and can be used as such.
Accordingly, I shy away from converting them to 8 bit, or doing any
adjustment of histogram. I suspect scan-from-disk output from an 8 bit
conversion would show degration of smooth tone.

It certainly would if you save it as linear output, since you are
failing to utilise, or take any advantage of, perceptual space -
consequently you are storing much more information in the highlights of
your scan than you can possibly ever see and, in the case of grainy
originals, ever use.

Convert the images to a gamma compensation of around 2.2, which has an
equivalent dynamic range of more than 16-bit linear with equal spacing
of the levels throughout the range. Stretch the histogram to achieve
maximum contrast without clipping any data and then reduce this to
8-bits. The resultant file will contain all of the image information
that can be perceived, especially from a grainy original.

If you insist on using Vuescan for certain editing functions, tell it to
use the embedded colorspace profile when reading the file from disk.

Scanning raw in linear space is simply wasting resources - indeed the
only reason the data is in linear space to begin with is because it is
sourced from a CCD scanner!

 

[Mendel Leisk]

This link will get you (I hope) to my Photo.net page, where I've posted
two examples of the stuff I'm scanning. One at 5400, one downsampled to
4000. They're in the folder "Downsampled etcetera"

http://www.photo.net/shared/community-member?user_id=517754

I believe you HAVE to be a logged-in member for the link to show the
pics. May work, may not.

Please, any and all, have a look if you can find it, and I would
appreciate opinions. I think the grainy/soft slides I'm doing don't
really suffer from the bicubic downsample, but any comments welcome.
Thanks.

 

[simplicity]

Why do you consider that to be important?

Scan from disk is no more than an alternative input into Vuescan - in
fact it is just another name for the File Open function in any image
editing application. Vuescan is a data capture application, and
performs those editing functions that it provides with less capability
than a proper image editing application - as you have now discovered.

I have been mystified by this "Vuescan scan-from-disk workflow" for
quite awhile. Thank you and Don for clarifying this.

On a slightly different topic. Is a digital camera's "camera raw"
similar in this regard? That is, the camera captures an image (like a
scanner's hw captures a scan), and then the image (or scan) is processed
by sw (i.e. camera raw) for better results?

 

[Don]

Downsample was not an easy decision. I know full well, 'bout the time
I finish, hard drive and dvd size will be doubled.

I'm scanning somewhere north of 2000 slides. With downsample to
4000dpi (and crop of black edges), they'll JUST about fit on my
(mirrored) 250gig drives. That is 16 bit rgb tiffs. They are
grainy/handheld/mixed quality. I want to do them all, don't want to
cull. At 4000dpi, it will take about 50 dvd's for the raws.

I looked long and hard at the difference, and decided I could live
with the reduced zoomability. Usually, zoomed into 100% at 5400dpi,
I'm looking at some very well defined grain on a softening subject.

I know exactly what you're going through because I'm battling the same
demons... :-(

I have about 1250 slides, another 750 negatives and about 1000 photos
for a total of some 3000 images. I also don't want to cull but want to
preserve everything. I was also hoping to fit it all on one drive but
gave up since I got the LS-50 and went from 2700 dpi to 4000 dpi
native resolution.

To make matter worse, since virtually all those slides are Kodachromes
and I twin-scan, I would really like to keep both scans (nominal and
shadow-boosted) as digital negatives. That alone doubles the slide
requirements... Sigh...

Add to that all the documents I will be scanning on my flatbed
(letters, etc) and I think that nice aluminum CD box for 300 DVDs I
saw in the store the other day may just be the ticket!

You're right about the drives, though. When I started, the largest
drive at my local shop was 160 GB, but I have since got a 250 GB as
well. In the meantime dual layer DVD burners are standard and I hear
that in Japan "Blue Ray" DVDs for HDTV have already been introduced
with ~30 GB capacity, or thereabouts.

Don.

 

[Kennedy McEwen]

I have been mystified by this "Vuescan scan-from-disk workflow" for
quite awhile. Thank you and Don for clarifying this.

On a slightly different topic. Is a digital camera's "camera raw"
similar in this regard? That is, the camera captures an image (like a
scanner's hw captures a scan), and then the image (or scan) is processed
by sw (i.e. camera raw) for better results?

It depends on the camera what is actually output as "raw".

The situation is slightly different with digital cameras in any case,
because they do not capture true rgb for each pixel, but interleave the
primary colours in a Bayer matrix. So each pixel only captures either
red or green or blue (plus a little of the other colours due to filter
leakages).

The camera software is then responsible for converting this into rgb at
each pixel using some fairly complex interpolation schemes. Some of the
better ones actually use the leaked sensitivity to adjacent colours as
part of the colour correction matrix in that interpolation, so that you
actually get more resolution out of the camera than a perfect rgbg Bayer
matrix would nominally permit. So you need some of that software to
actually get a raw image that you can store and subsequently process in
the first place.

Other cameras use a proprietary raw format so that they simply store the
Bayer data, and this can only be read by their proprietary software
because it needs to convert that data into a standard form before
further processing can be implemented, either in an editor or the
manufacturer's application.