22. August 2007
I know that LaserSoft Imaging normally produces high-class software, but when I saw the new Multi Exposure (ME) utility advertised on their site, I thought: "Well, OK it can't really be that good...again one of these 'amazing' new features..".
But I was wrong.
I downloaded the trial version and the first results was actually exactly as good as the examples from the site(!)
I just got myself a new and faster scanner!
The SilverFast software from LaserSoft Imaging is the de-facto standard for scanner software. With every new version supporting an even larger range of scanners - from the very high-end to the lower end scanners. The program pretty much covers and supports what you would expect from such a professional piece of software, including full color-management.
You'll find my introduction article here, which covers the (for me) most important features.
I will not give you a full review of the software here - many others have done that over the years, so I suggest that you follow the links below to read more comprehensive reviews.
What really caught my attention - and the subject of this article - is the new Multi Exposure (ME) function of the SilverFast V.6.5. A function that promises to increase the dynamic range of the scan.
I'm testing on my Minolta Dimage Scan Multi PRO - let's see how it performs!
Well, we've had Multi Sampling (MS) so far; a mechanism for layering multiple scans to eliminate noise from the CCD in dense areas. Some scanners supports this feature in hardware and it really can improve the dynamic range and reveal more shadow details.
However, the new Multi Exposure (ME) options seems to use a slightly different approach by scanning only twice, but with different exposure-times (?) The algorithm is patent-pending by LaserSoft Imaging and I'm not aware of the exact inner workings.
Getting started with ME.
Installing and setting up the SilverFast Ai version 6.5.0r5 on Windows XP was easy and trouble free. Calibrating using the IT8 target ran smoothly (see my former article for details on this subject).
The Multi Exposure option is then selected via this button:
Please observe that you must either select ME or multi sampling (which aren't the same, read above!) not both.
Almost any number of tests can be done when this kind of new software appears. You can test on positives, negatives, different film stock, any number of different software settings, dark shadow areas, light areas, with or without noise reduction etc.
I have to limit myself a bit, so I've decided (well, as usual) to limit myself to test using my normal workflow. This means that I'm testing on medium format positive films more precisely Fuji Provia and Fuji Velvia film stock.
I seldom use noise reduction (ICE or software), but I sometimes use it and will therefore also include it in this test.
I "assume" that methods for enhancing dynamic range and looking into dense areas of film is only really important for positives, slides more than for scanning negatives, but this is beyond the scope of this test.
In order to not just let SilverFast "compete against itself", I will compare results to the Minolta scanning software (Dimage Scan Utility) that came bundled with my Minolta Dimage Scan Multi PRO.
Conclusions are drawn based on several scans and much larger area than the small examples shown here.
This picture contains very dark and dense areas.
Fuji Provia (RDPIII)
It's Fuji Provia (RDPIII) in very low light. Not big art, but really good for testing.
Scans are all made at maximum resolution (for medium format scans on Minolta DSMPro): 3200dpi.
I'm showing a small detail (marked with a red square above) and I've brightened up heavily (changed the gamma using Levels Adjustments) to see shadow details. Efforts has been made to make the scans appear similar to ease comparison of noise and shadow details (Levels Adjustments again), making the colors rather unnatural. However I find that this setting reveals noise very clearly.
Minolta 1x pass
SilverFast 1x pass
Nearly identical scans. Differences maybe comes down to differences in the color profiles.
(Then why bother using SilverFast? Well, the workflow using SilverFast is very different from the Minolta software. SilverFast has build in features for calibration and profiling and the result requires far less processing afterwards. But that's a different story.)
And now Multi Exposure (ME) and Multi Sampling (MS):
Minolta 16xMS
SilverFast 2xME
On both scans noise is greatly reduced and more details are revealed. In this example the 16x multi sampling from the Minolta software contains ever so slightly more details and has a little bit less noise than the SilverFast ME scan.
Interesting result. When do we have a break even between ME and MS?
Let's take another examples - a Velvia 50 ISO slide, very dense in some areas and scan using only the SilverFast software, but scanning at 8xMS, 16xMS and 2xME.
Fuji Velvia (RVP)
SilverFast 16xMS
SilverFast 8xMS
SilverFast 2xME
In this scan, 16xMS and 2xME are nearly indistinguishable, but in this case the 2xME is leading with a very small margin. All though very close, I see a few more details and a bit less noise in the 2xME scan. On top of that I see vertical 1 pixel-wide green lines running through the 16xMS scan - they were on the Minolta scan as well, so it's produced by my scanner. No idea why, don't think I've seen it before.
The 8xMS is clearly left behind and shows a lot more noise and less details.
Can the 16x Minolta scan be better than the SilverFast scan, or is the difference here related to the film stock? Let's see a 16xMS Minolta software scan again:
Minolta 16xMS
It's very hard to tell the exact difference here. There are differences, maybe related to differences in the color profile. I think the Minolta is slightly better again, but I doubt there are any real image detail differences.
Scanning times
Scanning times becomes important here, because multi-scanning takes time!
SilverFast 2xME : 8:46 min.
SilverFast 16xMS : 18:25 min.
SilverFast 8xMS : 9:19 min.
In comparison:
Minolta 1 pass : 1:14 min.
Minolta 16 pass : 18+ min.
Dust removal (ICE) and ME
I will not make a test of how good (or bad) the ICE function is as a dust-removal option. I just want to test IF it works together with ME.
SilverFast 2xME and ICE
Unfortunately small artifacts are now seen, which can become a problem. I've haven't yet managed to produce a scan without these artifacts.
SilverFast has build in dust removal software, which in my experience works quite well, but at the expense of image details.
UPDATE [8.October 2007] : LaserSoft Imaging has made an update to solve the ICE problem. I'll be reviewing as soon as possible.
Glass or no glass film holders?
Although I have managed to produce a good looking scan using the glass-less holder, I normally get a "double exposure effect" with two not perfectly aligned scans.
Mid-scan calibration
It can happen that the Minolta scanner chooses to re-calibrate between two ME scans also resulting in "double exposure" images; two scans that are not perfectly aligned.
Not sure how to avoid this and what exactly triggers this, but it only occurs once in a while. Maybe one can re-calibrate the scanner before every scan to be sure that it won't occur mid-process, however this is quite tedious.
"Small-scan" artifacts
When scanning a small portion of a picture using ME one can get strange artifacts:
SilverFast 2xME artifacts
This never occurred on larger crops though.
I find it interesting to test against the Minolta software, because although the Minolta software has it's limitations, doesn't support calibration etc. I am (was) of the opinion, that if you just get the "raw" data from the scanner, you can't get more. Everything on top of that is just after scan processing. The only "hardware settings" on the Minolta scanner is cropping and focusing.
But, I'm very wrong here! The results shows that it really pays off to spend a little bit of extra time using either the ME or the MS option while scanning.
Given that the ME function produces an image hardly distinguishable from the 16xMS at less than half the time(!), I must say that the function is very usable and frankly quite impressive.
If your scanner does not support multi-sampling and you have to let the software scan several times and then layer the images, the time-differences could be much larger.
The ME function is not flawless as I've demonstrated, but the problems can be avoided and I'm now using ME as part of my regular workflow.
However, one whish would be to get rid of the artifacts when using ICE.
• SilverFast.com LaserSoft Imagings site.
• SilverFast Ai 6 - for Minolta Dimage Scan Multi PRO My former article about SilverFast.
• Minolta Dimage Scan Multi Pro - Review My review of the Dimage Scan Multi PRO film scanner.
|
|
|
Preface