Descreening. Please advise.

[Gregory C]

I don't have any professional equipment so I have to guess the screen density of printed matter by trial and error.

I usually begin with one of the presets that SF provides; eg, 133 lpi for magazines; and do a preview moir?. I then increase the lpi and 'update' the moir? to see if there is any improvement.

what I frequently find though; and what I find curious; is that if I take the best looking screen at standard densities (ie, 85 - 175 lpi) and then double it, the result is much better. for example, I just scanned a magazine article. with a Screen of 133 lpi, the descreening looked good, but at 250 lpi, the descreening looked GREAT!

could someone please explain why?


my frame settings were:
Scale 100%
Output resolution 225 dpi

scanning on a Microtek ScanMaker 5700 (FireWire)


by the way, I have to say that I'm really impressed with the output I'm getting from SF lately. I'm pretty sure that the Negafix profiles have been vastly improved (for SF Ai NikonM), and I have to wonder if SF has been improved internally in other non-documented (ie History) ways.

SF rocks (in spite of the UI and JM bugs I have reported of late) ;-)

regards
Gregory

[LSI_Belli]

Dear Gregory,

first of all, thanks for your contribution to this forum.

1. You can also "measure" the screen density by making a high resolution scan and count the number of patterns per inch

2. The scan resolution is internally set to the double of the pattern resolution. This enables to get the best results. So, the user cannot directly control the resolution.

It is strange that you get better results with a screen parameter higher than requested.

Here is an example of a raster of approx. 120 pattern/inch, and two scans with the screen parameter set to 120lpi and 240lpi. In this case, a parameter of 120lpi gives the best comprimise between detail accuracy and raster removal.

See the example below:

<a href="/img/forum/screenshotdescreening.jpg" target=_blank>Picture</a>

[Gregory C]

You can also "measure" the screen density by making a high resolution scan and count the number of patterns per inch

please explain further.

do you mean?
1. scan at 600 dpi
2. save the scan to disk
3. open the image file in an application like GraphicConverter
4. count the patterns

are the patterns that easy to see and count?

[LSI_Belli]

That's exactly what I mean. The hires scan should be done without descreening In the example I gave you, I measured the pattern periodicity by making a scan at 1200dpi (a smaller resolution would have been enough) and counting the number of patterns in a scan line (here approx. 9 per 90 pixels -> 120 per 1200pixels -> 120 "lines" per inch).

Thomas

[Gregory C]

I have just been working with Descreening again, this time scanning photos from a book for a relative.

experimentation with different lpi settings showed that the image was too blurry at 85 lpi, too moir?'d at 175 lpi, 200 lpi but very good at 133 lpi and 266 lpi.

what I am finding yet again is that setting the descreen density to double the original screen is getting better definition. at 133 lpi. the image contains no moir? patterns but is a little soft. at 266 lpi, there is a slight moir? pattern but far better definition. even turning on USM for the 133 lpi does not increase its definition to that of the 266 lpi image.

an added bonus of using the higher density is that I can save the image with a higher dpi setting. at 133 lpi and 400 dpi, SF automatically limits the scale setting to 65%. at 266 lpi and 400 dpi, I can save the image with a scale of 100%. the only problem is that 266 lpi/400 dpi requires a high-res scan which can take much longer to complete than a 133 lpi/400 dpi scan.

I have embedded three images here for comparison. it is a curious result because conventionally, I would not try to get more information than the original image contains; ie, images printed at 133 lpi should be scanned at no more than 133 lpi.

133 lpi with USM


266 lpi with USM


266 lpi without USM


regards
Gregory