Moire and Photoshop...

[DARPA]

This thread is coming here after the...ahem..."gentle nudge" from Al Gebra..

I originally posted this observation in my scan set thread...

Speaking of Moire, I found a couple of previously scanned but not adjusted images on my hard drive. Must have been for some test purpose I don't remember. But here's the thing, I started playing around with those in Photoshop and was downsizing them to 72 DPI from 300 DPI and using various pixel size ratios and I was seeing moires show up in Photoshop depending on what size I made the image. I have never seen that happen before where a moire popped up in Photoshop. Other times I had a mild moire with one un-adjusted image that got much worse if I shrank it to size A but dissappeared if I shrank it just a little smaller to size B but then came back again if I shrank it even further to size C (starting over from scratch each time of course). Can anyone shed any light on this. I scanned 5,000 images over five years and I never came across the particular moire phenomenon. I always thought moires only occurred because of print angle issues with a scanner and that once it's scanne in there, you're stuck with it. I'm totally unfamiliar with this peek-a-boo, now you see it/now you don't/now you see it again moire.

This eventually got a response from louiscar to which I will quote parts of and respond to here...

BTW. Many moons ago I discovered that litho type (dot prints) which were prone to patterning were better if I scanned much larger than the intended size. Sizing down in Photoshop reduced if not gets rid of the patterning effect. In some cases I've also noticed that this can also get rid of the moire patterning. However, if you have a technique to reduce the dots then the moire will disappear, they are of course key to why it happens in the first place.

First let me give the specs on my scanner. It's an HP Scanjet G4010. Given that I wasn't making a full time return to scanning I didn't want to spend a lot on a scanner and this one has proven to be a more reliable scanner than the one I did all my original mag scans 10 years ago. That one not only produced traditional moire patterns, which I could compensate for in Photoshop, but would occasionally create a weird swirling effect for certain mag pages, usually images that weren't full page or nearly full page, that you couldn't get rid of in Photoshop but had to rescan the pages at various odd angles until it would disappear. This new scanner so far has not produced an image with the latter effect.

I always scan at 300 DPI and 200% size. This wasn't so much because I was trying to avoid moire but because I would be able to end up with a substantially larger image after downsizing to 72 DPI (I would first downsize to 72 DPI but not change the pixel size and then lower the pixel size to get a larger image) but also because I would color adjust and use Levels on the image before I did any adjusting of the DPI and size as this would result in a more overall uniform histogram with no spikes (the downsizing would kill off the spikes).

I think though that you may also be describing a different problem and it's occured to me in the past that different graphic programs do interpolate differently when zooming in and out. Make sure you always view your sized pic at 100% otherwise you may just be seeing moire caused by the screen res vs image pixels vs onthefly interpolation of PS.

No this isn't a zoom issue. It's a problem in Photoshop's ability with downsizing images. And it's not a problem I used to have to contend with. Photoshop, and we're talking version 6.0 as that was what I was using back then, used to be able to get rid of the moire if you shrunk it far enough. But I'm finding that CS4 doesn't rub out moire the way 6.0 did, it preserves it in a cyclical way.

Say I have a scanned image at 2000x1000 with a 300 DPI. Moire exists at that size. Given the size of the original image, in the past using 6.0 my target aspect ratio when all was said and done would be about 500x450 with 72 DPI. And usually the moire would be gone.

But now it isn't gone. It's still there. Photoshop preserved the effect. But the twist is if I tried ending up with a different size setting, say 600x550, the moire could be worse or better or gone completely. I find that I have to keep trying different sizes for the final image to find the size that shows the least moire or no moire. We're talking a difference of +/- 5-10 pixels in dimensions - 610x560 produces more moire, 600x550 produces less moire, 590x540 produces a relatively moire free image, 580x530 produces more moire again. And so on. It's cyclical. You have to find the dimension in the cycle that produces the least moire/no moire.

And because it is cyclical, there are several possible "sweet spots" you can pick from. This can be good or bad. It can be good because I can now get final "clean" images which are bigger than I could have gotten 10 years ago. Whereas 10 years ago I couldn't get higher than 500x450 for one image, I could now get 600x550 or 700x650, depending on the image. A nice improvement.

But it's bad because there is no single formula that works for every single image. It's a crap-shoot and you have to constantly guess a size for a starting point and then work up or down in size to find the correct size that's moire free.

You may also find this article helpful:

http://desktoppub.about.com/gi/o.htm.../basics06.html

This article did give me a few ideas, particularly the blurring of the image prior to downsizing, which is something I haven't tried before. I'll have to give it a shot.

[louiscar]

Originally Posted by DARPA

First let me give the specs on my scanner. It's an HP Scanjet G4010.

So presumably no de-screen filter or something similar in the software options?

Originally Posted by DARPA

I always scan at 300 DPI and 200% size. This wasn't so much because I was trying to avoid moire but because I would be able to end up with a substantially larger image after downsizing to 72 DPI (I would first downsize to 72 DPI but not change the pixel size and then lower the pixel size to get a larger image) but also because I would color adjust and use Levels on the image before I did any adjusting of the DPI and size as this would result in a more overall uniform histogram with no spikes (the downsizing would kill off the spikes).

Ok this is where I differ. Even 10 years ago I scanned at a min of 600dpi. It was painful then but now not so bad, I'll even go to 800dpi or higher if necessary. I also don't have a size option as well (in the software - dpi setting only), it comes out at whatever size 800dpi on that area makes.

This makes a huge image of course but it gives me the leeway to size right down usually to less than 50%. After which he evidence of any kind of litho dot pattern is totally gone.

Let me just illustrate that because I just uploaded some old scans I did of Cara Lott:



Rest of them are here: http://www.vintage-erotica-forum.com...cara-lott.html
Post #102.

That was not a great scanner in those days and yes it's still big but you don't see the dots, they're gone therefore now I could size that to any size I like and not get any moire whatsoever.

As for screen / print dpi changes, I usually don't bother. As you know in PS you can change from 300dpi to 72dpi without resizing the image. You check the box below to not resample. However, if you are not doing that you are using the 72dpi to reduce the size as well and will get whatever size it equates to. I prefer to put the target size in directly, eg. 800 x 600 or whatever. 72dpi is usually for screen and 300dpi for printing generally they can be changed independently of actual size later.

Originally Posted by DARPA

No this isn't a zoom issue. It's a problem in Photoshop's ability with downsizing images. And it's not a problem I used to have to contend with.

Ok it's an interesting one and as I don't do a lot of scanning these days I haven't noticed it. I do have a project coming up so I'll look at that and see although I'm using CS3 right now. I am doing old photos (non porn) which some do have a texture which causes similar problems. My goal is to preserve a enough detail to archive these photos as once they're gone they're gone. Of course any kind of patterning or moire is unacceptable to me so I'll scan much bigger to get rid of it and retain at least the same detail that was in the original.

What I'd do is try the same thing in some other packages to see if you get the same cyclic effect. Perhaps Paintshop pro which is my second choice for doing some quick and dirty things. Let's see if it's an algorythm.

I'd also be willing to confirm for you if this is a CS4 bug perhaps. If you'd like to send me a troublesome example I can try this on CS3 to see if this has the same effect.

Originally Posted by DARPA

But it's bad because there is no single formula that works for every single image. It's a crap-shoot and you have to constantly guess a size for a starting point and then work up or down in size to find the correct size that's moire free.

And perhaps another test here is get to your sweet spot and then check out other sizes in paintshop pro or other package. Did the effect come back? if not size slightly smaller so you at least re-write the jpg and then put into Photoshop and see if the problem still exists.

Originally Posted by DARPA

This article did give me a few ideas, particularly the blurring of the image prior to downsizing, which is something I haven't tried before. I'll have to give it a shot.

I do think that scanning very high first and reducing is a good method because you still retain the sharpness. The more pixels you have to play with the less reason to blur to get rid of the dots, the sizing is doing most of the work for you. The downside is time - it will increase the time each pic takes to get on disc but then when you are looking at long time preservation it's worth considering.

Although there are colour shifts due to my original scanner I am happy with the sharpness of the Cara images and most of the scans I did in those days, I've not tried this on my current Canon Lide 500 but it does seem a reasonable scanner. Each scanner's CCD will also be part of this equation, a subject I'm not that familiar with, eg how they differ now days to then.

I'd be interested though in finding out if CS4 has a particular problem or perhaps other later versions after 6.

[DARPA]

Originally Posted by louiscar

Ok this is where I differ. Even 10 years ago I scanned at a min of 600dpi. It was painful then but now not so bad, I'll even go to 800dpi or higher if necessary.

It was too painful for me then an it's still too painful for me now. Back then it was taking me 1-2 hours to scan in a set at 300 DPI. That's just the scanning, not the post scanning adjustment part. If the end result was a (mostly) non-moire set of images that was good enough for me. I didn't need to go to a higher DPI.

I find that is still true now though I'm still getting acquainted with what this scanner can do. It now takes me under 45 minutes to get a set scanned in. If I can do an end run around this photoshop moire wierdness by blurring the image at 300 DPI that will speed up the process even more for me.

I'm not trying for the "perfect image". Good enough more often than not is acceptable for me.

As for screen / print dpi changes, I usually don't bother. As you know in PS you can change from 300dpi to 72dpi without resizing the image.

Previously I didn't have to go this route...I could just downsize the DPI to 72 while resizing the image. But this time around I am downsizing without resizing as the first step. Then I'm resizing as a second step.

Ok it's an interesting one and as I don't do a lot of scanning these days I haven't noticed it. I do have a project coming up so I'll look at that and see although I'm using CS3 right now.

I'm going to scan another set tonight so I'll save one series of an image at various sizes to show this effect. Of course, this is a moot point if blurring eliminates the moire. But I'm not convinced this is necessarily a bug in CS4. It could have been a flaw in 6.0 that didn't precisely dither the image as well as CS4 does when resizing. In other words, a better dither could preserve the moire in a way that 6.0 couldn't. Which would mean I was accidentally achieving success with 6.0 because it wasn't precise enough.

I do think that scanning very high first and reducing is a good method because you still retain the sharpness. The more pixels you have to play with the less reason to blur to get rid of the dots.

This is true. But I'm not convinced yet that I need to go above 300 DPI to achieve acceptable results. In my time from 1998-2004 I only encountered a handful of mags where the dot printing was so egregious, or the image being scanned was so small, that you couldn't dither the dots away starting at 300 DPI.

The one area of my scanning which is a continual work in progress as well as a bane of my existence has to do with the touchy subject of color correction. I think I'll bust out a separate topic on this subject to stimulate discussion and feedback.

[louiscar]

Originally Posted by DARPA

It was too painful for me then an it's still too painful for me now. Back then it was taking me 1-2 hours to scan in a set at 300 DPI. That's just the scanning, not the post scanning adjustment part. If the end result was a (mostly) non-moire set of images that was good enough for me. I didn't need to go to a higher DPI.

This is true. But I'm not convinced yet that I need to go above 300 DPI to achieve acceptable results. In my time from 1998-2004 I only encountered a handful of mags where the dot printing was so egregious, or the image being scanned was so small, that you couldn't dither the dots away starting at 300 DPI.

Ok I understand the issues and of course the work / time factor vs acceptability is a pretty important part of the decision making. I sympathise fully with your view particularly when you are doing so much of it. It will take your life over unless you can find a quick path.

Originally Posted by DARPA

The one area of my scanning which is a continual work in progress as well as a bane of my existence has to do with the touchy subject of color correction. I think I'll bust out a separate topic on this subject to stimulate discussion and feedback.

As a photographer I know that problem well too. if you've got a white and a black in the picture it can be very easy but alas this doesn't always happen.

[DARPA]

Ok the Gaussian blur really made a huge difference. Despeckle doesn't seem to do much (I do despeckle first). But I'm still using it.

Since the moire problem is no longer a problem I'm not too inclined now to do a comparison set of pictures to show what Photoshop does with moire when you downsize to various sizes.

[louiscar]

Originally Posted by DARPA

Ok the Gaussian blur really made a huge difference. Despeckle doesn't seem to do much (I do despeckle first). But I'm still using it.

If blur is acceptable to you then experiment also with the median filter you may find it better at retaining sharpness where it matters but it's a case of suck it and see.

[DARPA]

Unsharpen Mask seems to do the trick for retaining sharpness...

[muler]

Well, after reading all of that, I guess I must inform you that I scan using my HP Scanjet 3200C. This is a parallel scanner, old timers must have heared of it, it was a very low end scanner at that time, I got this one for free, had PSU problems, I replaced one capasitor in there and it works like a charm!

This one takes about 15 minutes to scan a page at 400 dpi, I use this setting because the default of the scanner is 150 dpi and this produces moire many times.

After completing the scan, I save it as JPG using highest quality. Then I use Irfanview to resize the image at maybe 1000 pixels horizontaly using the Lanczos resample filter and adding sharpening after resizing.

After all of that, I post the image over here.

I guess this must be the reason that I do not post that many images...

[tuffy]

Originally Posted by muler

Well, after reading all of that, I guess I must inform you that I scan using my HP Scanjet 3200C. This is a parallel scanner, old timers must have heared of it, it was a very low end scanner at that time, I got this one for free, had PSU problems, I replaced one capasitor in there and it works like a charm!

This one takes about 15 minutes to scan a page at 400 dpi, I use this setting because the default of the scanner is 150 dpi and this produces moire many times.

When you choose 400dpi, you get a very large file without any new information (no new info above 200dpi). It's better to use 300dpi or 200dpi.

After completing the scan, I save it as JPG using highest quality. Then I use Irfanview to resize the image at maybe 1000 pixels horizontaly using the Lanczos resample filter and adding sharpening after resizing.

You use jpeg compression two times, it's not good. You can save the scanned picture in .TIFF format, a TIFF file using lossless compression (or none) may be edited and re-saved without losing image quality. The most important thing is to use the Descreen funtion if you want to avoid the moire.

Let me illustrate this:

1. your original scan

2. how it looks like on my screen (full of moire)

3. your scan with Descreen ON: (created in Photoshop)

<-- click to enlarge

I hope you can see the difference. Please observe the file size too (482kB vs 281kB).

After all of that, I post the image over here.

I guess this must be the reason that I do not post that many images...

Please test the above advice and ask more questions if something is not clear.

Tuffy

[muler]

Thanks for the info.

I tested scanning at 200 dpi and 300 dpi, the aquired image may be very large (for a JPG at 100% quality) but my joke of a scanner does a better job higher than 300 dpi. At 200 dpi there is moire you wouldn't believe. The scanned pic looks even better at 600 dpi but I am too bored to wait that long.

Here is my masterpiece of a scanner...

http://www2.shopping.com/xPO-Hewlett...-ScanJet-3200C

It is correct that I should not save a JPG twice, actually the second time I use 90% quality to end up with a smaller file.

No way of using anything like photoshop, my computer cannot handle such an application, it has the age of my joke scanner (if not more).

Once I had scanned and saved as TIFF, produced very large files indeed, then made a multipage image file with these. I have to try it again maybe, altough doing that would propably mean that I would have to erase the original TIFF files, due to hard disk space constraints.

The image you descreened looks like a bit blurred on my side, so I will check if the blur filter (of Irfanview) does any good in mixing the dots together.

Anyway, I can do a lot of tests on a scanned file, but cannot do many scan tests because of the 15 minutes per page, also while using the scanner, some other specific apps on my computer do not work because of some IRQ or (more likely) DMA conflict.

Makes me wonder why I bother to scan at all, but it looks right to try to give something back to this great forum.