What am I doing wrong here?

[jamiewakeham]

Hi all

Got my slides of my alpine expedition back from the lab recently. Borrowed a mate's DualScan IV and put the first few through.

The scans really don't seem to be very sharp. I don't really know what to expect - this is the first time I've scanned my own slides - but if I look at the images any closer that full-screen sized, they seem pretty soft. In comparison to, say, my girlfriend's 4MP digicam, they're appreciably a lot softer.

Before anyone says it, yes the slides themselves are sharp... 🙄

I've attached a crop from one of the scans. It's from Velvia exposed in a Canonet QL-17, hyperfocal at f11 or f8, scanned at 3200dpi in 16-bit colour, 16 passes, with the dust tool activated but the other scanner options turned off. I used the noise->dust&scratches tool in Photoshop Elements, corrected a slight blue colour cast, and used 150% USM at radius 3 and threshold 1.

It shouldn't be that soft, should it? Those branches really are razor sharp in the original, as is the line between the mountain and the sky.

Any advice?

Cheers,
Jamie

 

[taffer]

How small is the crop re the full frame size ?

I always have to apply a post-scan sharpening to anything that comes out of my 3170 to make it as close to the original as possible. What I recall is having very bad results from applying dust&scratches, as the loss in sharpness was dramatic.

 

[jamiewakeham]

Hi Taffer

It's perhaps 10% of the width and 5% of the height (I've taken a landscape crop out of a portrait frame).

So I should start by not applying the dust & sctraches routine? I've got another example straight out iof the scanner... hang on...

Jamie

 

[daveozzz]

That's definitely too soft for that sort of resolution... To me the softness looks like postprocessing gone a bit wrong. What was the scan like before you did the photoshop stuff?

Do you really need the dust&scratches filter when you used ICE on the scanner...?

I'm not familiar with that scanner but does it definitely support multiscanning? I know from experience with my Nikon Coolscan V multiscanning can intoduce some softness since the scanner isn't able to exactly position the film the same every scan.

What I'd do is start with a 1 pass scan only at the max resolution with ICE on it's lightest setting and everything else off. Any chance you could show us the image at this point?
Whatever, stay away from the PS dust&scratch filter - it just blurs the image - I virtually never get any evidence of dust after an ICE scan and if I do I remove by hand with the healing brush.

Also try leaving off the sharpening until you're sure the scan is OK. I never touch images with sharpening until the very last step after I've resized for my final print or web page or whatever (and your settings seemed quite heavy... I very rarely go higher than a radius of 1).

Hope yo have better luck your next attempt!
Cheers,
Dave.

 

[jamiewakeham]

Right - this is a (slightly larger) crop from a scan that I've done nothing to in PS at all. As before, it had 16 passes (the Dualscan has this as a option and indicates that more passes = better quality, but I'll certainly try single pass) and the scanner's own ICE turned on.

It seems less soft, but still nothing compared to the output of the gf's digicam.

Cheers,
Jamie

 

[daveozzz]

Hard to say without knowing exactly what size crop it is.. easiest way is to go to Actual Pixels or 100% zoom and cropping a section of that.
This doesn't look disasterous... still looks really dirty considering ICE was on - looks like what I end up with dirtwise ICE turned off.
Would you say it's definitely softer than the original then?
I'd certainly try 1 pass to see how that does. And sharpening will help a lot but what I was saying the the post before is to go easy on it and only use it once you've resized for your final use since you'll need different settings for different resolutions.

I've attached a 100% crop of one of mine you can use as a comparison. (1 pass @ 4000dpi on a Coolscan V, light ICE, no sharpening - taken with velvia 100) - I don't think they're too different.

 

[dmr]

If you do a maximum resolution scan, can you clearly see the detail of the film grain?

Does the grain appear to be in focus all over the frame, or does it "soften" toward the edges or toward the center of the frame?

Oh, if it's the Scan Dual IV, it doesn't have ICE, but the auto-dustbrush, which I have not found to be particularly useful.

 

[ch1]

I use a Nikon 5000 coolscan and always scan in RAW (or NEF as Nikon calls it). Huge files (65mb or so) is the "price" for doing that - but it will give you the greatest possible resolution to work PS with.

What level of resolution are you scanning in?

Also, I agree, it would be better to see the un-PS'ed scan to see if its the scanner or you post-scanning processing.

 

[payasam]

I see two small dark spots, one to either side of the top of the helmet. On the helmet itself, and only in the shadow on it, there are eleven larger circles. Two of these, most clearly the one nearest the logo on the front of the helmet, have small white dots at the edge. Might all this say anything?

 

[jamiewakeham]

Dave,

how do I tell exactly what size crop I've carried out (sorry - not good at this digital stuff. I think I'm a bit analogue...)? I can tell you it's pretty much 10% of the area each way, or I can look at the quoted size of the JPEG? 800x600, btw.

I can barely discern any grain in there, DMR; what there is is too small for me to be able to tell if it's in focus across the frame. But you're right, it is 'dustbrush' and not ICE. I'd assumed they were equivalent, but I guess I'm wrong on that! I'll try turning it off. It certainly doesn't seem to be much good, given that the second crop I've posted is with it turned on!

Copakje_Ham: yep, I was getting RAW scans of 83MB at 3200dpi. I was scanning in 16-bit - I assume this is better, even though PS immediately asks me if it can convert back to 8-bit. I also assume that agreeing to that is the right thing to do?

What would happen if I tried to attach the 80MB RAW scan for you guys to look at? Would it be useful? Would RFF be unhappy about it (I guess so)? I could go back and delete it later today.


Thank you all for your help!

Jamie

 

[daveozzz]

Yip scanning at 16bit is best but I think you only really see an advantage if you do any image PS manipulations in 16bit mode since there are more colour choices for PS and you don't get the comb-effect histrograms you do if you manipulate an 8 bit image (which can end up as posterisation). All images have to be in 8bit for printing so you should convert to 8bit once you're done any manipulations.

Not sure how you tell in retrospect exactly what size a crop was but sounds like you've taken a fairly small sample. If you zoom into the same image the last crop was from so that the same parts of the crop fill the PS canvas then what % does it say on the title bar after the "@" symbol?

I'd guess attaching a raw scan would be considered bad... maybe convert to 8bit and save as a jpeg quality setting 5 would be good enough for a look.

 

[dmr]

I can barely discern any grain in there, DMR; what there is is too small for me to be able to tell if it's in focus across the frame. But you're right, it is 'dustbrush' and not ICE. I'd assumed they were equivalent, but I guess I'm wrong on that! I'll try turning it off. It certainly doesn't seem to be much good, given that the second crop I've posted is with it turned on!

Oh, before I forget, there is one trap that I've fallen into. When you first start it up, it sets something like 800 dpi resolution for everything. When you set it to something higher, it's for that frame only, not all of them. Your second scan may be at the lower res. That threw me at first.

I'm attaching an example of a blow-up at maximum res showing the grain and showing the focus at the edge and at the center of the frame. The one at the edge is at the top of a "portrait" mode frame. Notice how it softens as it approaches the upper edge. The center grain is sharp and crisp. This is from a 1975 vintage Tri-X, so it shows the grain quite well, but you should get the point. At a normal size there is no apparent sharpness fall-off on the edges.

If there's any siginficant bend in the negative, you can set the focus. I've found that doing autofocus on an off-center point of high detail works quite well, but often this isn't needed.

I've really found that auto dust brush to be a placebo at best. It didn't seem to have any effect on the old dirty negatives at all, so I quit trying it.

In any case, I've never had any difficulty getting very sharp scans from this thing.

 

[Rob Holland]

Jamie:

I looked at your second pic in Photoshop and it does indeed look way too soft for a Velvia scan. I have the SDIV and get excellent results from it.

As someone has already mentioned, the KM auto-dust brush is less than useful. I suggest not using it. Makes the scans very soft AND isn't all that effective.

16X resampling is way overkill unless you have a lot of deep shadow area. Try it with none and see how that does. Even with high contrast velveeta, I've rarely need more than 4X. Now that I think about it, the 16X may be part of your softening problem. It's averaging each pixel 16 times and may be softening the edges.

Not much need to scan 16-bit if you don't have 16-bit post process software, such as a later version of Photoshop.

Sharpening settings (USM) depend on the intended use of the scan. For full-res prints a good starting point is amount 85, radius 1.0 and threshold 0. If it's a bit soft, just reapeat the unsharp mask once more with the same setting. For web viewing, try amount 400, radius 0.3 and threshold 0. If it gets the 'halos', reduce the amount to around 200 and see what it looks like. All sharpening should be done after final resize, of course.

Hope this helps. Looks like you had an awesome trip!

Later,
Rob

 

[jamiewakeham]

Thank you all so much for your help.

I'm off home in a few minutes; I'll report back in the morning with some single-pass scans without the dust-buster thang turned on!

I've attached a medium-quality JPEG of that second scan; as a warning to those on dial-up, it's a little shy of 950KB.

EDIT - nope, can't get it to upload, even at setting 1! I assume I'll have to pull the resolution way down to get it to fit, which rather defeats the object of you guys being able to look at it at full res... no worries. I'll have another go at the scanning tonight.

And yes, it was one hell of a trip! I'll tell the stories when I get all the scans done. Suffice to say that the QL-17 was the sole totally reliable camera on the trip... my Pentax SPII was doing well until the mirror froze to the bottom of the pentaprism 😱

Night, all.
Jamie

 

[jamiewakeham]

Hi again, all.

This is a similar crop from a scan I made last night (in a hurry, hence perhaps a little dustier) with only one pass and the dust buster turned off. Still at 16 bit depth and 3200dpi.

Seems to me not to look very different to the first example at all, though it does seem to respond to USM better.

I'm still not sure I'm getting everything I could from my slides, though. In fact, I'm still not sure I'd have got rather better images than these from my GF's digicam (if I'd been able to stop it from freezing, anway 😉 ) so I still wonder if I've got something not right.

Don't suppose anyone would let me email them the scan I took last night, and just tell me if they would expect more sharpness or not from a scan? It's 83MB, I'm afraid.

Cheers,
Jamie