Silverfast AI

[KevinBE]

Does anyone have the AI version of Silverfast? I am really getting into the LE version that came with my Epson 2450. I really like what I can do with it. Their approach to making the tonal corrections in the scan is really good. I also like the descreen filter, it really works well. Only problem with the descreen is that it limits you to 15 Meg file size. That along with the lack of scanner color profile support are the only two things I don't like about the LE version. I was wondering if the AI version addresses these two problems.

[Stephen M]

Kevin - I currently do not do much scanning (our drum operator takes care of most of that), but some stuff such as descreens and art which is not flexible are flatbed scanned by me using SilverFast AI, v4 I think (older version).

Yes, SilverFast is favoured by many and has many good options. Infact it is very deep for a scanner app and can rival Photoshop in the words that need to be written to get the best from this app.

Whenever this subject comes up, it is also well worth looking into VueScan and perhaps also ScanPrepPro as two other common popular third party scanner drivers (for flatbed or film scanners).

Try a search for info on all three, but the talk is really BIG on VueScan considering it's features and update policy etc. I have no association with any scanner software maker - of the above three I can only comment on SilverFast 4.

As for the full versions abilities, yes the descreen does work and the only limit is scanning at 100% size or smaller magnification (since you don't want to enlarge dots that you are tyring to minimize <g>).

I scan at 100% mag with the original at a 15 deg angle for a 45 deg mono halftone or scan with the original at 45 deg angle for a full colour image (but it can vary). This helps avoid halftone pattern and scanner moire. I also scan at 600 ppi instead of 300 ppi and also use scanner descreen. This is a slow scan but well worth it. Then I crop angle back the scan from the scan angle to the true horizontal angle in Photoshop.

Then I run my special action on the file, which takes a while and uses a few steps but the end results are usually worth it. The final file is near 100% size at 300 ppi.

Insed of using the action, a quicker method would just be to sample down at this point to the final desired size/resolution (single step, 33/66% steps or incremental steps of 10% etc). Perhaps a quick hue/chroma clean up and or other quicker tricks and a sharpen etc.

It is rare that I am happy with just a descreen from scanner software with no other cleanup work.

As for SilverFast custom colour calibration - have not used it so I can't comment. As flatbed scanning is not critical for us, our setup is less than 'ideal' for flatbed scans when it comes to colour accuracy in input (I am left to colour correct by softproof and digiproof).

Regards,

Stephen Marsh.

[KevinBE]

Thanks Stephen. I really appreciate your indebth responce. I think I will take a look at Viewscan just to see if it will do what I want. It is sure a lot cheaper than Silverfast AI but it will have big shoes to fill.

The color management I want is for a ICC aware program that will scan using the scanner profile I have created in Profile Prism and convert it to Adobe RGB in the output, like the twain driver does.

Thanks again for your input. You have given me a lot of useful information.

[Stephen M]

OK - Firstly here are some links to moire:

http://members.ozemail.com.au/~binar...V_links.html#M (scroll down a bit).

Many of the hue/chroma moire can be handled with methods like:

http://www.ledet.com/margulis/ACT_po...-blurring.html

http://members.ozemail.com.au/~binar...V_links.html#L

Dan Margulis also has some good ideas on descreening, with the angle scans mentioned above being some of it. His v6 book had a whole chapter on moire and descreening.

I have attached a sample of some scans I did a few years ago for a client (crop of a client supplied halftone print for a performer doing a show for the client).

600 ppi optical resolution scan of reflective art, halftone print:

Left image - original with 90 degree regular scan with no scanner descreen. Centre image - descreen scan with 90 degree scan angle (regular scan). Right image - 45 degree angle scan with no descreen crop rotated back to true angle in Photoshop.

Top row is the original 1:1 view of the RGB composite channel. Lower row is the same image viewed at half size. The left image is the original to compare against, the right is the final result after I run my descreen action which filters, blends and reduces the final pixel count down to near half size (approx).

Keep in mind this is a 25% quality JPG of a screen shot!!!

If you download this pic you will notice the difference between the channels with the various scan methods.

Some quick points to consider:

* If your descreening ability is currently limited - you may be able to scan without descreen but using other combinations of methods which provide near the same result, which may or may not be as quick.

* You can combine scanner descreening with other methods to gain from both techniques.

* The angle of the scan can dramatically affect the luminosity moire and the less visible colour channel moire, which can be addressed by filtering with a fade/blend to color mode or by filtering the AB of LAB.

* Scanning with descreen can be slow, but so can angle scanning as more area is captured before the angle crop in Photoshop.

* Without scanner descreen you can often do similar things if you know the tricks, or combine the tricks with scanner descreen to go even further.

* Grain and noise reduction methods are also good for cleaning up descreens after you have tried the above methods:

http://members.ozemail.com.au/~binar...V_links.html#G

That's enough for now, descreen can be as simple or complex as your output and requirements dictate.

Stephen Marsh.

[KevinBE]

Thanks Stephen, it will take me a while to go through all this but it will be worth it. This is very helpful! I just finished a project to restore a graduation picture of my mother-in-law. Like a lot of portraits made about 50 to 60 years ago the paper used had a raised micro-demple texture to it which the scanner picked up on real well. The Lasersoft LE program descreen filter did a great job of removing the Morie' but , the 15 meg file limit hurt the output file dpi. The twain driver is limited to 600 dpi with descreen on but the descreen filter just doesn't work very well. Your links will give me more options as to how to deal with this problem in the future.

Thanks again for your hard work getting this information together for me.

[Stephen M]

Kevin - the good news is that I have more techniques for you to play with...the bad news is that I have more techniques for you to play with!

Flatbed scans of rough surface textures...

* Scan once as normal, then try a 180 degree flip and another scan (ensure straight scan edges etc). Rotate and align each scan as layers in Photoshop using difference mode and opacitites etc. Some rotation may be required but is not ideal, things should match up but you need some luck!

When aligned blend the layer in lighten or darken mode, depending on the artifact. Also add blurs/medians or other noise reduction attempts and blends as required etc.

* Another approach is to try a second scan that is 90 degrees offsest to the original scan, then rotate that back and align and blend as above.

* Instead of a second scan, use whatever noise reduction methods you like and blend those into the original with lighten or darken or other modes and opacities.

* The add and subtract blend modes in Apply Image offer possibilities that do not exist with layers...

Flatbed lighting casts long shadows in one direction, acquring a second scan from a different direction will light the surface defects differently...then blending the two together or using a difference mask or whatever you can make use of this often unwanted 'feature' of flatbeds and surface grain/texture.

In this case, two wrongs do make a right. <g>

Stephen Marsh.

[roger_ele]

Hi, I have the latest version of Silverfast and I just downloaded Vuescan and compared a badly faded color original scanned on both. I am not a scanning expert, but with this one test Silverfast automatically handled the original much better than VueScan. We use Silverfast all the time and like it alot, if you can afford it I think it is worth the gamble. I beleive you can download a trial version to test for yourself...
Roger

[chris h]

Exhaustive review of Silverfast 6 here..


http://www.computer-darkroom.com/sfa...verfastv61.htm



It would appear that versions of Silverfast are specific to individual scanners.

[Ed_L]

If you really are interested in Vuescan, I posted a thread yesterday on available tutorials for it. This should give you a great indication of it's capabilities.

Ed

[KevinBE]

Wow. You guys have given me exactly what I needed.

I looked at Viewscan, it seems to be a very powerful scanning program. I didn't see the color management I want but I could have overlooked it. This program appears to be similar in power and capabilities of Qimage Pro in that it is hard to fully appreciate it's capabilities without an investment in time learning what it can do.

I think I will upgrade to Silverfast AI. It isn't a lot of money, $119.00, and it seems to be a very powerful scanning tool. 75% of all my work comes from scanned images so it makes since to have the tools to make this part of the process easier and better.

Stephen, your tips are invaluable. I'm going to start using them on my next project. I think that there will be a lot of times when the Morie' in the originals I will work with will need a better technique than the scanner can provide. I have noticed that the scanner filters leave the pictures very soft. So if that is the only way to reduce the Morie' it will be something I have to deal with. I think that some of the techniques you gave will have a better chance of doing the job with minimal damage to the image.

Thanks again guys.

[KevinBE]

Thanks Ed I'ill go through the materal on Viewscan. I think that there is a lot more to this program than is imediately evident.