Whats wrong with my scanning technique? - compare these two images

[ted_smith]

Hi

Please compare the two images below.

The first is a 'standard' (as opposed to pro) 4Mb scan of a Fuji Velvia 100 35mm exposure done by my pro lab that they burned to CD, along with the rest of the film.

The second is a 600dpi 24-bit RGB scan of the same transparency using my own Nikon LS-2000 film scanner with a 'auto-levels' colour balance selected during the scan process. I've done nothing else with it, either in the scan software or afterwards, and I accepted the defaults of the scanning software more or less, which is VueScan, version 8.5.01 (www.hamrick.com).

I realise I could correct it with sharpening etc, but the two scans are so different, and I expect the pro lab are only using default values too, so where I am I going wrong? Surely it's not just down to the fact they will be using a much better scanner, or is there really that much of a difference?

Where am I going wrong? What are the pro lab doing that I am not?

Ted

 

[PVia]

Your highlights are blown, and the color correction is non-existent. After previewing your scan, set the shadow and highlight levels so as not to clip them. After that, once you're in Photoshop you can color correct the image.

Plus, 600 dpi for a 35mm frame comes nowhere near what you need for a 4mb file...

Plus, it looks like somewhere in your workflow, there is dirty glass...

 

[Garry Madlung]

Scanning

Hi Ted

I noticed your post on apug and thought I would reply hear. I can't give all the answers, but I can give some of my experience. I had a quick look at your site (nice work). You seem to know what you're doing. I might not be able to add anything.

As you mentioned, there is a big difference between the lab scanners and consumer or prosumer scanners. Your pro lab is probably scanning your trannies on a drumscanner or at least an Imacon. I own an Epson 4990 but rarely use it.

My Epson is supposed to be very good, and I can get excellent prints from it. The colour is still different from my Imacon and drum scanning. I'm not sure it's possible to get the same colour from different scanners. Different software? colour sensitivity? I'm not sure.

I do scan my Velvia slides flat, without adjusting curves or colour. It's best to keep the information intact before it's brought into Ps. Adjust those when you are in Ps. Perhaps, if you are selecting 'landscape' colour balance, the software is interpreting everything to the blue and green side. My only suggestion might be to just do a straight scan, making sure only that you are not clipping either end of the levels graph.

My Epson can remove dust, but I am told that it is best to do this in Ps to avoid spoiling data. Incidentally, drum scanning and Imacon scanning do not remove dust. Dust removal could just be a convenience invented for consumers. I only take my newly developed slides out of the sleeves moments before placing them in the Imacon scanner to attract as little dust as possible.

I also do not add any sharpening in the scan process.

As well, the drum/Imacon or scan is so sharp that you need to "despeckle" (photoshop>filters>noise>despeckle) to get rid of gritty quality. Maybe you know all this; excuse me, I don't want to be too didactic. I'm just sharing my experience.

The cost of scanning can be steep. I have a lab in Toronto that will rent an Imacon for $25CDN/hour (if I prepay). The regular rate is $50/hour. Drum scans are charge by the MB and the price adds up if you are scanning several slides. I still love Velvia though, so I'm selective about what I scan.

Garry

 

[Garry Madlung]

Excuse that long reply. Here's the abbreviated version:

- Make a straight scan. Don't use 'landscape' colour balance.
- Don't adjust curves.
- Don't clip histogram (as mentioned).

 

[Tim Gray]

I've gotten this on scans that were really dense (kodachrome). When I went to adjust curves to make things look 'right', I got a similar effect. I think it's actually due to flaring from the highlights contaminating the shadow areas. It became more noticeable in my case because I had to crank on the curves to get 'proper' exposure (just trying to match what my eye could see in the slide).

I don't know what to do about it though...

 

[Bruce Watson]

The lab you used almost certainly did *not* use the defaults for their scanner and software. They almost certainly did create an ICC profile for their scanner and Velvia. Additionally, they probably did some tuning to make sure that the film they scan is always in the scanner's plane of focus so they'd get nice crisp scans.

As you have found, scanning isn't as easy as it looks. It's not a push-the-button-get-perfect-results kind of deal. Like just about anything worth doing well, there are learning curves to climb and that takes practice. The more you do the better you get. As long as you are trying to improve of course.

As to what you did wrong, three things come immediately to mind. You clipped both your black and your white points. That's why your shadows are empty and your highlights are blown. You used auto levels, which is why your colors are off. And your focus seems to be off for some reason, I have no idea why as I have zero experience with a Nikon LS-2000. It also looks as if some part of your scanner's optical path might be dirty -- the upper right corner seems to be "blooming" which looks like flare of some kind.

That said, there's nothing wrong with the scanner you have. Work with it, learn the software, and scan, and scan, and scan some more and you'll keep getting better and better at this.

 

[jd callow]

Bruce seems have hit all the notes.

Here is what I see:
Auto adjustment is clipping too much at either end
colour balance is off (Auto adjustment issue)
contrast is too high (as above)
scan is out of focus (can't help you)

 

[pellicle]

Hi Ted

Firstly, nice picture.

Secondly can I ask you to not place such large images on the screen amid the text? It causes everyone's posts to wrap all dreadfully. I see you hosted them off your own site, so (even with this clumsy interface) you could have put some 500pixel wide (at the most) scaled images and then had that linked to something larger. Like this



I first inserted an image (clicking on the image icon above) then high lighted that text and and inserted a link (pasting in the url to the bigger image).

that just makes reading things less of a scroll back and forth ... pardon me

now, on to answering your questions.

There are two issues

1: your scanner needs cleaning

I have a similar generation nikon (the LS-20E) which I liked quite a lot when I bought it. I found that my scans had exactly that same 'soft' flare which is all over yours, and I thought it looked like gummy optics on an enlarger.

I pulled it apart and cleaned it and it made an enormous difference. Sorry but back then I was not as careful a documenter as I am now. Instead I've found a link which shows how to to it with your scanner (even better as the LS-20 is different). Here.

Some advice though ... My friend had the LS-2000 which I thought had much nicer D-max ability than mine, but the ICE sucked. A few years after that I had the opportunity to test a LS-30 (on which ICE also sucked) but then after that the LS-IV (which didn't burn down or sink into the swamp) inherited the best scanning of scanning sharply and better DMax and ICE which worked fantastically (LS-4000, LS-V and LS 5000 are only better yet).

To wit here is a comparison of the same bit of Provia scanned to what I could get out of the LS-20 (dust cleaned by hand) and the LS-IV (but not much else done except resize and put into sRGB for display)

Now some things will be come apparent here: The cleanness of this image shows that this is my post cleaning LS-20 and that the contrast is different. Look at how much more shadow detail I could get with the newer scanner!

But there is no free lunch, so while you can get into the shadows more it means that the (now wider) range of the image is compressed into the same 255 levels (making it seem like the LS-20 is punchier straight out of the scanner). This can of course be tweaked in photoshop with some use of curves.

So, what I'm suggesting to you is to consider:

• do you clean your LS-2000 and have better contrast but still an inferior scanner
• or do you ebay it and get a nicer LS-IV or better

I've recently bought an LS-IV for not much on ebay ;-)


2:

People have identified the issue of clipping the dark and light points of the image. I have a blog article (using the Epson but you could equally use the Nikon) which talks about the black and white points and their significance to noise / high light and shadow details here.

I think that you need to more carefully set the black and white points of your scan (the point either side of the histogram as seen in the above article) a little wider than the software would recommend. I often use "auto exposure" on the scanner then go in and widen them a little more by hand.

Clearly this opens up noise (and on the LS-2000 green channel noise is present as specks in the black), I adjust this with curves in photoshop, so as to keep the main part of the image unchanged and gently trim off the edges (where the noise and blow outs are). Like this:

Personally I have never (ever) been comfortable with vuescan (and without a word of blasphemy, God knows I've tried), perhaps its useful in batch scanning. But because I scan one image and work with it I continue to use Nikon scan or Epson scan.

If you can, perform as much as you can in the scanner software (not colour balancing though) and deliver to Photoshop the best raw material you can. Consider that your scanner software is like your raw converter for your digital camera.

Like others have said, scanning is rather an artform in itself. A sound technical understanding of what is going on will help remove the hit and miss, as will attention to what you're doing and trying to improve. Feel free to PM me or chat about this topic if you wish (and assuming I have time) I'd be happy to help.

HTH

 

[ctscanner]

Ted,

Here are some thoughts about your question.

There is a gentleman named Wayne Fulton, who has a website you might want to visit: www.scantips.com. He pretty much covers the basics of scanning in a pretty clear manner. There used to be a book available thru the site called 'A few scanning tips', it's a pretty good investment if it is still available.

Vuescan is a good scanning package, but, so is Lasersoft Silverfast, which in my opinion is easier to work with and has more intuitive controls. It is more expensive, and will only support the one scanner, unlike Vuescan which supports a host of machines. I have been very pleased with the results from Silverfast.

I stay away from 'Auto' anything during the scan process - reason being that I prefer to set the image parameters as I feel they should be, not as the software would dictate.

Try scanning at the maximum resolution your scanner supports, and then adjust the file size once you know what you are going to do with the image. This approach might improve sharpness, but should allow you to capture most, if not all, of the detail in the image.

Sort of general I know, but perhaps something you can use.

George