Scanning experiences: Moirée patterns and white point

Discussion in 'Tips, Tutorials & Tools' started by Leif Oh, May 4, 2005.

  1. Leif Oh

    Leif Oh Member

    Trying to get better control over scanned working copies, I've learnt two fairly important things recently, namely how to avoid disturbing "moirée" patterns, and how to bring out the colours in full clarity by establishing a correct "white point".

    Avoid "photo" setting when scanning!

    Like many others I use the ubiquitous Epson scanner and inkjet printer. Understanding the different settings in the "Twain" menu (what you enter when you scan images) has always been a mystery for me. By chance (and some bitter experience) I learned a bit more the other night.

    I wanted to be ambitious and make a really good scan. Not only did I decide to go to 300 dpi (from my usual 150), but I also chose the "photo" setting for both the kind of document I printed, and for the "target" (that is, the printer), thinking that this was the most "advanced" setting.

    Big mistake. Not only did the files turn out to be absolutely huge, since I scanned at 206 percent (from 1/33 to 1/16). This doubling in scale means a quadrupling in filesize at any resolution (which is unavoidable if you enlarge). If you double the resolution as well, the result is no less than 16 times larger files than if you scan at the original scale and lower resolution. (A standard size sheet in 1/33 turned out to be around 50 MEGs at 1/16 and 300 dpi, which is a lot.)

    Worse still, there was a distinct moirée pattern in the scan (if you don't know the term, I'll show you what it looks like below). First I thought that this was only visible on the screen, but a test print clearly showed up traces of moirée as well.

    Pondering this, my only conclusion is that if you choose the setting for "Colour document" (like I've done before, more or less by chance), some clever software is engaged, whereby moirée patterns are eliminated or reduced (since printed images are made up of a mesh or grid of little dots, or pixels, which is what creates moirée patterns when superimposing the new mesh of pixels created by scanning; while photos contain no such original mesh of pixels).

    In trying to be ambitious, I had made the mistake of inadvertenty disengaging this clever piece of software, by choosing the settings for "Colour photo" and "Epson Stylus Photo", instead of "Colour document" and "Epson Stylus Printer Fine".

    Having a theory of the cause of the mistake was all well and good but it needed to be confirmed. So I scanned the same original with all imaginable combinations of settings. Here are the results:


    [Above:] To the left is the control scan, made with "Document" settings for both type of document and target (printer). I already knew that choosing "Photo" options for both document and target/printer would produce moirée, so I tested the two remaining options. To the right, I've retained the "Document" setting for type of document, but changed target to "Photo". The resulting moirée pattern is clearly visible. (There is some moirée pattern also in the control scan, but not much and even less so in the full size original; check for yourself by clicking in the image).

    Now, this difference could be because the target/printer should be set to "Stylus Fine" instead of "Photo", so I changed that:


    [Above:] To the left, still, the control scan with "Document" settings for both scanner and printer. To the right, the scan was made with the target/printer changed to "Stylus fine", while the document type retained the "Photo" option. Still very evident moirée pattern.

    Conclusion and bottom line: If you use an Epson scanner, use the settings "Color Document", and "Stylus Printer Fine". Avoid anything that implies that you're scanning a photo, because then you will in all likelihood get moirée patterns. (I suspect that there are similar settings for other scanners.)

    As for resolution, I'm reverting to my usual 150 dpi, which is fine by me, particularly since I enlarge to more than twice the original size already in the scanner. I think the limiting factor then is print quality and not scanner resolution. But I'm ready to listen to advice from more experienced people.

    Getting the white point right in scanning

    When you scan an original in order to enlarge it (or make a safety, or building copy) the background inevitably turns out at some other shade than white. I've always got rid of this with the help of the "magic wand" tool, which is available in any graphics program, thinking that losing all this smudge would save me some precious ink when printing.

    Not until the other night did it strike me that a different shade than white of the background also necessarily meant that all other colours where slightly "off". How to fix this?

    Searching through the Photoshop menues I found something called "Levels" (translation from my Swedish version), which I so far only had used for correcting photo exposures. Looking more closely I now found, in the "Levels" dialogue box, a number of small pipettes designated as "black point", "grey point", and "white point", respectively.

    Eureka! Clicking the "white point" pipette, I could select a point in the scanned original, which I wanted to be white (I chose an empty space in the middle of the scan), and, hey presto, everything - not least important, the colours in the parts themselves - became just a little bit brighter, and the background absolutely white.

    So here's my newly learned practice of treating scanned documents in Photoshop (or any good graphics program):


    [Above:] Here's the scanned original, with the "Levels" dialogue box brought up. You may notice how the background is slightly golden grey.


    [Above:] Choosing the pipette to the right ("White point"), and sampling an empty space in the middle of the scan, makes the whole scan brighter, and the background pure white. Notice in particular the difference in clarity of colours.


    [Above:] And now you can finish up by getting rid of also all the background white by using the magic wand. (If nothing else, this will make it easier two superimpose two partial scans when recreating a scanned original.)

    Since these adjustments are made in the graphics program, older scans are not lost. Even if you have removed the background with the help of the magic wand, there is bound to be some white area in the parts themselves untouched by the wand (such as e.g. glueing tabs), which you can use as reference for sampling the correct white point.

    Main thing learned: Scanning an original and just leaving it at that (or simply removing the background) will produce a slightly too dark version. Establishing a correct white point in a graphics program will bring out the colours in full clarity.

  2. cadwal

    cadwal Member

    A very nice article Leif.

    Regarding that white point correction. Are you not supposed to use the same white point settings for all "related" scans? Making sure that the scans are made at the same time using a properly warmed up scanner in order to get an even effect on the entire model. Otherwise colours might "drift" among different images after the adjustment.

    Looking at the Gimp it wants the user to pick both a white point and a black point. I suppose that make sense if the black areas suffer from the same problem as the white ones.

    Edit: I forgot to say that I have also noticed that I should select "document with graphs" instead of "something photo" on my HP/Canon combination of scanner/printer to minimize image artifacts. I think that the algorithms used when selecting "photo" for scan&print cannot handle images that are as "regular" (lots of solid colours etc.) in appearance as a normal papermodel.
  3. Gecko23

    Gecko23 Member

    The 'clever' bit of software isn't too clever, its just doing some averaging. You can get the same results on a scanner that doesn't do this by scanning at a higher resolution, doing some careful blurring (photoshop's 'smartblur' is a good start), then reducing the resolution.

    Adjusting the white balance is a good idea, as is replacing the background with a solid color. Most every file format will compress the image better this way.

    Personally, I stick with 300 DPI or higher. Even if you eventually intend to print it at 150 DPI, the higher the resolution when you are editing, the better. Every time you reduce the image size, you lose information that you can't get back.
  4. jrts

    jrts Active Member

    Hi Leif

    Some very good tip's there, dam good write up.

    I never go lower than 300dpi, seem's to work well for me.

    Thanks again

  5. k5083

    k5083 Member

    I suggest sticking with 300dpi or higher and using a sophisticated noise and moire reduction program to clean up the scan. I use Noiseware Community Edition which has the advantage of being freeware. It analyzes patterns in your scan to guess the frequency of the moire patterns and blur accordingly while leaving other details sharp. Results can be amazing, sometimes an improvement on the original published model, especially with older Modelarz, Modelcard, and similar that had rather gross halftoning. Get here:

    (Go down to bottom of page for the freebie Community Edition.)

    As for white points, I always set those and the black points as well. The two rules I observe are (1) keep track of the settings and apply the same ones to all pages of the kit, and (2) be sure to change only the levels and not the color balance; the latter can be done manually, but you don't want the computer trying to color balance your scan as if it were a photo.

    August (a lurker who will get around to saying a proper hello one of these days)
  6. Leif Oh

    Leif Oh Member

    Thanks guys for some very valuable advice. I agree with the 300 dpi level of resolution. In my special case, however, I think I am on that level, since I enlarge the models to more than twice their original size, and scan at 150 dpi. The pixel content of my scans is probably the same as yours (a regular Halinski sheet scans in at around 12 meg).

    At that magnification, I deem the print quality to be the limiting factor of quality, and therefore feel pretty confident about my 150 dpi. If you were to scale down one of my scans, I think the quality would be equal to a 300 dpi scan at the orginal scale.

    If I were to scan a model at its original scale, though, I would definitely follow your advice.

  7. Jim Nunn

    Jim Nunn Member


    You need to put this in the Tips folder.
    This is an excellent example of practical experience put in to words I can understand. I’ll defiantly give this a try when I get back home.

    Jim Nunn
  8. Gil

    Gil Active Member


    Amongst other things I was scanning in GPM's F8F-1 Bearcat over the last several weeks..., the results seemed to be all over the map, the color map that is. I took a break from this activity as it was going absolutely nowhere when I read your thread. I let it sink in for a week and a time period which could be devoted to gaining some positive direction in this effort. There are several items which should be added to the list which will also help.

    Your white point was a great tip as it solved some issues I was having. Setting the bottom level up consistently also resolved the bottom "stretch" issue that was bedeviling the process. I actually ended up rescanning at 360 dpi using "Color Documents" scanning mode. Printing was also accomplished using the manually set "Graphic" mode. The Navy Blue was not a perfect match as the gradients and noise of the original just do not make the scan translation. Staying away from all "Photo" modes in both scanning and printing seems to be key in preserving maximum original content in the final print copy. In other words stay away from algorithms.

    Additional items that should be added to the list for optimum image scanning and image transfer are:

    o Do not set the scanner in bright light. Cover with a piece of cardboard or a blanket during the scan process.

    o When scanning a model do it all in one session using the same settings. This preserves page to page color fidelity.

    o Find out what the recommended scan DPI should be for your printer. For the Canon i560 Bubblejet series 360 DPI is recommended.

    o Use scan and print categories that avoid the word "photo". Look for terms like Line Art and Document. Avoid anything that appears or looks like it is associated with an algorithm.

    o Don't try to achieve perfect color fidelity to the original. You will fail. The object is to obtain a multi page copy which maintains the same color from page to page and is fairly close to the original. Remember that human color memory is very poor without reference so differing colors next to one another will be painfully apparent.

    This is in addition to your original post which could form the bulk of a scanning tutorial.

    Best regards, Gil
  9. Leif Oh

    Leif Oh Member

    Great! This should be very helpful to everybody. It is obviously a ubiquitous problem, and not one recognized immediately even by experienced people.

    Good tip about covering the scanner (which I have seen before on the site somewhere, but forgotten). I have my printer in a pretty dark corner already, but find it imperative to put something heavy on to the lid, booth to prevent light from getting in, and to avoid distortion from slightly bent pages.

    Even so, the edge connected to the part "hanging out" from the scanner is invariably more smudged than the rest. That is usually discarded, however, in the process of combining the partial scans into a new, scanned, reproduction of the complete original sheet.


    PS. The problem of differing colours is very apparent in the Airacobra I just finished. If you have a look at the photos, you will immediately spot the difference between front and tail ends of the fuselage, as well as the wing fairings. I'm not sure that is due to the scan, however, since that was made pretty much at one go. The reasons more likely are: 1) printing at different occassions, and 2) different mode of treatment with clear acrylic.

    So, not only should you scan all parts at one go, you should also print them in one sequence. That's an awful lot of printing... (not to mention spare parts being printed as needed). - L.
  10. Gil

    Gil Active Member


    This is reminiscent of animation where all the color for a production run would be mixed up at the beginning and only that paint would be used for coloring the matte cels. Color changes would be kept to a minimum using this method. The same method should be adopted for building card models that have been scanned. Checking the printer ink prior to making a run is a good idea if not mandatory and a supply of ink cartridges from the same manufacturer is also suggested by this practice. One other item that should be added to the list is to clean and align the heads before running the print job. The completed print should have the surface sealed with matte acrylic, clear acrylic or lacquer implying that an adequate supply of the sealer be on hand to complete the present run and enough for future "rebuilds" as necessary. A quick and easy way to warm the spray can and it's contents that works with any canned aerosol product is to place the can under hot tap water agitating it to mix and warm the contents (wipe off any residual moisture before spraying). The contents warm in under a minute or so and result in a very fine spray mist.

  11. k5083

    k5083 Member

    One more tip for fine-tuning the color of scans. Print yourself some small stickers, all at the same time, each one consisting of blocks of pure red, blue, yellow, black, white, and medium grey. Affix one identical sticker to a blank area of each page of the original paper model you are scanning. (Use temporary adhesive if you don't want to deface your original model.) After applying all post-scanning corrections discussed in this thread, use your graphics software's eyedropper or color sample tool to analyze and compare each color of the scanned sticker on each page. Fine-tune colors separately until all are the same. Don't expect the scanned colors to be "pure" (i.e. the red block will not necessarily be r=255, g=0, b=0), just make them the same as each other.
  12. Gil

    Gil Active Member

    That's a nice idea. Do you keep a log of adjustments to each page making the same adjustments to the other pages? The reason for the question is the adjustments to one page might not be the same for another resulting in page to page color differential which if placed side by side on the model will be quite evident.

  13. gera

    gera Member

    Just to suggest to all what our friend "k5083" (?) :roll: suggested in his post.Download the program NOICEWARE (free)at
    and you will have a nice surprise. I downloaded it a few days ago and have had VERY GOOD results with my scans. I will use this program from now on..........Thanks "k5083" (????????? :? )
  14. k5083

    k5083 Member

    Thanks Gerardo. My new avatar may help explain my handle. :) Noiseware really is a gem.

    Gil, I do keep a log of the gross adjustments and make the same ones to each page of the model, but when it comes to the fine tuning, the goal is to make DIFFERENT adjustments to each page to correct for page-to-page variations in scanning that survive the various excellent ideas (dark room, warmed up scanner, etc.) discussed on this thread so far.

    The point of having the identical colored stickers on each page of the model is to have a color calibration reference on each page. If you can get the sticker colors on page 1, page 2, page 3, etc. all consistent in the scans, then the colors of the model parts should be consistent as well, and you will not have mismatched panels resulting from page-to-page scanning variations. Of course, if you have done a really good job of controlling the scanning environment and made the same basic adjustments to all pages, the colored stickers should just confirm that the pages all match and you don't need any fine tuning.

  15. Gil

    Gil Active Member


    The reference you mention is commonly called a "test coupon" and is used as an incoming quality control element on numerous entities in industry and now card modeling..., Keeping notes is a mandatory practice as not doing so leads to complete chaos.

    This thread is yielding great information!


    I'll be jumping over to the site you listed.

    Best, Gil
  16. Gil

    Gil Active Member


    After thinking about the "test coupon" awhile it became apparent that it is dependent on the paper that the coupon is printed on. The displayable gamut differs markedly with the paper type. Printing a CMYK and grayscale test coupon on the best paper type yields the theoretical color limits of the printer. Applying this to scanning then needs to take into account that some fidelity is lost in scanning. What we see on the screen is sort of in between the two which also handles the extent of the color gamut differently from either the scanner or printer..., hope you're still with me. Sampling the colors of the test coupon is the result of one, a printout (test coupon) two, a scanning process and three, sampling from the screen with the eyedropper tool. Now tweak the color channels to adjust the color toward what the original colors were. Print out the test pattern (best paper again). Repeat this sequence several times and see if the color converges or diverges. I have the distinct feeling it will end up somewhere in the pastels due to the white washout effect of the process. Interesting to see where it ends up...,

    Best, Gil
  17. Gil

    Gil Active Member


    Looked at Noiseware. Interesting plugin for Photoshop. It seems to handle the chore of "adjusting" photos automatically which can be done manually using the standard feature set. I'm somewhat leery of software that "learns" more about you as it is used..., How much success have you had using it to scan with? That's really what counts a successful scan that preserves resolution and color and essentially what we're all after.

    Best regards, Gil
  18. k5083

    k5083 Member


    On the test coupon (thanks for educating me on the jargon), I fully agree that the scan, and printouts of the scan, will vary from the original sticker. I don't actually use the stickers to try to match the scans or printouts to the original model. My use of them is much more limited: trying to get the colors of the different pages of my scans and printouts to match each other. I'm looking for consistency, not fidelity to the original. Test coupons of course can be used for both, but matching a printed-out scan to an original paper model is rather ambitious and I prefer to do it by eyeball. Actually, I don't generally pay much attention to whether my printout matches the original model because the truth is that the colors of almost all published paper models, even by the best makers, are pretty inaccurate. So if I'm scanning and printing, say, Halinski's Messerschmitt, I focus on getting the colors to look like those of a real Messerschmitt rather than replicating Halinski's colors.

    As for Noiseware, it does not scan, all it is is the intelligent averaging algorithm that you have been playing with. In principle this could be done using the Photoshop averaging tools but that is not just a chore, it takes a lot of skill, practice and talent. If you can match Noiseware's results manually you're a better Photoshopper than I am! I also was leery of the learning at first, especially because I use it for several very different applications (digital photos, scanned paper models, grain in scanned negatives) and wondered what it might make of all of this variety. However, after several weeks I have not noticed any difference in performance so I don't worry about it now.
  19. Gil

    Gil Active Member


    I can't think of a better way to use the test coupon method! Your explanation regarding color fidelity of published models is well taken and serves to put the process into a proper perspective.

    You mention that you use Noiseware for scanned paper models. Can you expand a little on its benefits?

    Best regards, Gil
  20. gera

    gera Member

    Hi Guys.....great thread, I have learned a few things here....

    Gil: Like 5083 says, Noiseware does not scan.
    My procedure for getting the best scanned immage follows the lines of 5083 :? .....
    1)I scan almost always at 150, sometimes I raise this and I really notice "nothing" except the scanner takes longer.
    2) I scan to Photoimpact, which I have been using for about three years, maybe more.
    Here I do what 5083 does, try to get the most real color for the plane. I check photos etc and a) I adjust Brightness and Contrast. Then I try to get the best possible color for the part and "write the values down". I use my fill tool to get the parts painted with each individual color. I use the same values for the "other" scanned pages so I get consistency through them all.
    When I thing everything looks ok, I do a "test" print to the paper I will be using to build the model. This is very importat since I found that "you always have to do some re-touch!!!!". The screen shows one thing and the paper another :x
    That´s it. Now that I have used Noiseware, tried it in al old Maly HS-123 and PE-2, and I can say its the best run I have had in years!!!!!.....I did full around with the controls a lot and wrote down the best settings. I think I will now simple use it on every scan and smile........ :lol: , before this program I hated to scan......really good.

Share This Page