How to remove the paper texture from a scan?

<p>I'm in the middle of touching up an old family photo. It's hung on

the wall for ages, and is on its last legs. It's a nice 11x14 print

on some sort of matte or slightly textured paper.</p>

 

<p>I scanned it on an Epson 1250 flatbed at 400 dpi for output, and

have just finished getting rid of dust and scratches. One thing

that's bugging me is the pattern of the matte paper that the scanner

has picked up. I'm used to scanning negs, not originals, so I'm not

familiar with this problem.</p>

 

<p>Does anyone know of a technique in photoshop to sample a pattern

and remove it from the image? Does this sort of thing exist? I'd

love to hear anyones thoughts. I'll attach the full image and a

cropped 100% chunk with the pattern if it helps anyone.</p>

 

<p>Cheers!</p><div></div>

And here's the full image:<div></div>

Why not just run it through a noise reduction filter with a sample taken from the pattern?

Try the 'despeckle' filter in photoshop. Be warned, it can destroy sharpness, too. I usually

despeckle and then fade.

Above advice is fine, but I am wondering why you think it will 'show' in you final output. Does it show in the original, and if so, does that bother you?

 

Copying small prints on stippled paper surfaces and enlarging them is a a problem because the texture enlarges with the image. In your case (I am guessing!) you will probably reduce the image, and therefore the surface texture. Try outputting a hardcopy sample to test what you see. You may be happy without despeckling which does soften the image.

Save it in jpeg format, 8 bit (which for this image should be more than enough) and try out the free version of Neat Image. It might do a passable job of cleaning things up while minimizing the loss of sharpness.

There's really no way to get rid of such an effect. It's modulation of the light, which is multiplication, not like dust, scratches, and noise, which are all addition and subtraction. And it's not a "pattern", it's just random, at a lower spatial frequency than most noise. A noise removal program will tear that image apart, without really fixing the noise.

 

Is an optical effect that's only annoying at a 100% crop of a 400dpi image really that much of a problem? Have you tried a sample print yet?

 

This is a perfect example of the GIGO principle - the Epson has a single lamp, and bumps and dips really cast shadows, creating the modulation. I scan stuff with a strong texture on a two lamp Microtek.

<blockquote>A noise removal program will tear that image apart, without really fixing the noise.</blockquote>

Good noise removal programs also consider low frequencies. This is what I got with Noiseware (and the difference amplified, for comparison). It is hard to tell whether the softening would be objectionable on the whole based on this sample, because it has very little detail to begin with.<div></div>

Hi

 

Some scanner programs have a "de-screen" option that can do quite a good job with some textured surfaces.

Can you post a link to the full sized scan? I (and others, I'm sure) can run it through NI (or another noise filtration) program and post the results. With reasonable filtration settings, I suspect NI will do a fine job on the image.

 

-Greg

Thanks to all for your thoughts - much appreciated.

 

John, and others - I don't find the texture on the original print unsightly, but it's the paper texture - on the 'outside'. When I print it on the dlab, it'll be on the 'inside'. Don't know if that makes sense, but it's tricky to put into words. (I'm an Italian - it'd make perfect sense if I could talk with my hands :)

 

Mind you, it might not be as obtrusive as I think once it's printed. I'll run off an 11x14 and see what I think. I don't want to sacrifice any sharpness by despeckling, or a similar method, because it's not very sharp to begin with. I may try what Emre has done on a seperate level and tweak the opacity to get a happy medium.

 

I won't bother linking an copy of the original, as I think it's about a 50mb greyscale flattened.

 

Again, thanks for your help.

Oh, and BTW, Joseph: interesting to know about this GIGO business. I'd be very keen in trying out the Microtek and see how it compares. As I said, I've not done any scans of original reflective images, and I'm sure I could have gotten a better initial scan to work with.

 

Cheers,

Another approach would be to copy the print photographically with a good macro lens (ideally flat field prime lens, such as Nikon Auto Micro Nikkor).

 

One doesn't lose as much making and scanning a good 35mm film copy at high resolution (3200 or 4000) as one does with a coarse reflective scan (400ppi).

 

Making a film copy would allow you to play with lighting types, lighting angles, and, if there' any gloss in the original, a polarizer.

... you might try some careful gaussian blur followed by careful unsharp masking on your existing scans.

 

...some images actually appear sharper overall after that process, even if tiny details suffer. Tiny details, such as frizzy hair, may not contribute as much to the look of "sharpness" as do some bigger details, such as eyes and hands, that might be improved with gaussian blur and unsharp masking.

I found this similar question on another page and added to it:

 

Don Gough , Aug 31, 2004; 10:18 p.m.

 

When the problem is just surface texture and not that the image is darker on the raised portions of the surface, I have reasonable results by scanning with the print in the "upright" position, then again upside down, one on one layer in Photoshop, the other in a second layer. Rotate the topmost layer as necessary and position it accurately using Blending/Difference, then use Darken for this top layer. Accurate positioning is the biggest problem, it may be that only the most critical portion of the image may be good - at least you'll get that part clean.

 

Don G.

 

***

More on the above...

 

Cheryl Levy , Sep 01, 2008; 12:21 p.m.

 

Don G's solution really helped a lot! It didn't get rid of everything, but the improvement was noticeable.

 

In it, he mentions "using Blending/Difference" after pasting the 2nd scan and lining it up (hint during line-up: go into the "layers" menu and change the opacity of the top layer so that you can see the bottom layer, then make it 100% again). I made sure the top layer was selected, then clicked on "Difference" in blend mode -- the whole screen turned black. So if Don G could explain what that selection does, that might help. HOWEVER, instead of clicking on "difference" I clicked on "Darken" and noticed a difference right away. Then you can close the box "hit "ok" and if you still see some speckle all over you can "move" the top image (make sure the move tool is selected and then just use your arrows). Keep an eye on things like eyes because you might make them too blurry. If you still need to adjust, say, a big area of hair, but don't want to further blur the eyes, use your lasso tool to select the area you want (feather the selection if you need to) and then use the move tool to adjust that area some more -- the eyes (or whatever other areas you DON'T want to further blur) will be left alone. Play around with it on a sample scan so that you can see what is happening. When you're done you can merge the layers. Thanks Don!

The reason the "paper texture" shows up is becuase of the lighting; its not flat with most all flatbeds. The light source; the scan bar and sensor are at a grazing angle; ie like the moons craters "pop out" more when its 1/4 illuminated; than when full. Here I use an older scanner that has a very long optrical path; its not a regular scanner at all.Thus with its flatter lighting the moon craters of textured paper are quashed; this scanner is integral with a 16K printer.

<p>Phtoshop CS4 has a Blur Texture under the Blur filters section. It does very good job removing texture , however it comes at a high price. overall softness.... <br>

When texture is bad , we have to learn to compromise. I just had a small job of a picture from 1943, the paper texture is about the most annoying I've seen , it showed as little perfect squres of high contrast all over the picture, even worse, over people's skin. I had to settle on the softness and then high pass sharpness. I did not like the results. </p>