Untitled Document

Greetings Color Theory,

I'm having major problems with moiré patterns and am hoping to find help here. I've uploaded three files to the Photo Albums directory in the album "Big_Bad_Moire".

I've spent a couple of days hacking the images by using techniques from Dan's book (by copy is ProPhotoshop ed. #4) and from various posts in this group. I have not succeeded and I'm hoping for a fellow member to help out of the woods.

Most solutions to moiré patterns generated from digital capture require blurring and blending. My problem is that the subject *is* a 2-D matrix of LEDs. Any blurring or softening of the display ruins the feature of the product: nice, sharp, bright images. If anyone has an idea to eliminate or radically reduce the moiré, please post.

My current solution is to either 1) use film, scan and pray there's no moiré when the image goes to halftone and 2) photograph the display from a distance that the frequency of the LEDs (19mm) is less than the frequency of the camera's sensor elements (7.2 microns).

Any suggestions are greatly appreciated.

Cheers ~

Chris Brown

___________________________________________________________________________

In the days we got moiré with scanners, we turned the scanned page by 30°.

Have you tried this approach? Turning the cam 0° and then turn the image back in PS.

Maybe it's a stupid idea, but just popped into my mind that this could work here...

Regards,

Werner Tschan

--

STUDIO LTD

Atelier für Fotografie

Altenbergstrasse 8

CH-3013 Bern

T: ++41(0)31 332 88 33

F: ++41 (0)31 331 62 42

M: ++41(0)79 227 02 19

U: www.studio-ltd.com

___________________________________________________________________________

Hi Chris,

First thing that came to mind was trying to shoot those things at night. Maybe the daylight hitting the panel is what is causing the moire. Also a change in the angle at which you are shooting might help as well as using a very slow shutter speed.

Please keep us posted and when you finally solve it let us know how you did it.

All the best!

Paco Marquez

___________________________________________________________________________

On Mar 13, 2007, at 3:28 PM, pacomarquezfoto wrote:

I agree.

yup.

Not as optimistic on this one, but easy to do. Also try extremes of aperture. I'd also try a polarizer - easy and cheap.

What camera body? What software? RAW? In-camera filtering or sharpening?

Interesting problem!

--Rich Wagner

___________________________________________________________________________

The moiré is not coming from a pattern clashing with halftone dots - it's burned into the image itself. In other words you can keep the LED texture and get rid of the moiré. If you can shoot the objects in a few different ways and see what works best I'd go with that.

But otherwise you'll just have to fix the moiré in Photoshop. And IMO when the moiré is actually part of the image the solution is not simply to blur, but to color correct. (Although, come to think of it, a surface blur may help with the LED texture if you can get a tight mask).

Question - how many images are we talking about?

--

J Walton

___________________________________________________________________________

I must photograph these displays at all times of day, high noon, morning, dusk, & evening. This shows the brightness of the LEDs, and the ability of the display software to adjust brightness for dusk and night.

This does not help. I never photograph the matrix straight on, but always from the side (for pictoral reasons). I will post an example tomorrow.

Camera: 1DsM2 capturing raw images

Lens: 45mm TS-E, 24mm TS-E

Polarizers if it helps the sky, but never in the dusk or evening

Software: Capture One Pro, no USM

Output to RGB (camera's colorspace)

Thanks, you guys.

Chris Brown

___________________________________________________________________________

Chris,

Have you looked at Dan's "Photoshop LAB Color" book? There is a technique for dealing with moire that looks like it will work with a little fiddling in your situation. I tried blending the a channel into the L and can minimize the moire (on the red background), however the low resolution and JPEG artifacts make it difficult to see how effective it is.

Another alternative if you do a lot of this type of work is to look at a "re-screener" filter. http: //www.caprockdev.com/antimoire.htm I have a couple and they have been of use with medium format digital backs (no anti-aliasing filter like most DSLRs).

Darren Bernaerdt

___________________________________________________________________________

On Mar 13, 2007, at 4:28 PM, Chris Brown wrote:

I'm not looking to affect the sky, but to possibly kill the moire at the source - the same way that you use a polarizer for glare. As it's easy to do, I'd still try it. It's not completely clear what is causing the moire, but it's a much cleaner fix if you can do it in camera rather than in PS.

Shooting at night would also give a clue as to whether the problem is cause by the LEDs themselves, or perhaps by sunlight interacting with the plastic of the LEDs.

Good luck!

--Rich

___________________________________________________________________________

Moire is aliasing, and can only be prevented, not removed afterwards, because it's indistinguishable from real low-frequency content.

Are you saying that the moire appears in the original raw file from the camera? What body? The Canons I've got have strong enough anti-alias filters that you never see moire, but I know some cameras don't. If that's the issue, the best you can do is defocus a little when you take the original shot. That's not a perfect solution, but it may help.

Ciao, Paul D. DeRocco

___________________________________________________________________________

There was a time...in the far distant past...where cameras used something called 'film'. These cameras can be had real cheaply. They will seldom be preferred to a digital camera...but if this is something you will be shooting over and over again you might want to get one. Never, ever, saw a moiré with a film camera.

Gene Palmiter

Chris Brown

___________________________________________________________________________

The moire may appear when you scan the film. It's an alias of the pattern in the image, when sampled at the spatial frequency of the sensor. The only solution is to make sure that the image has no spatial frequencies that high, which takes a diffuser. Defocusing can reduce it, but not eliminate it.

Ciao, Paul D. DeRocco

___________________________________________________________________________

I didn't realise this is scanned film. Paul's post brings something back to mind - can it not be eliminated by selecting a lower scan resolution?

Mark Segal

___________________________________________________________________________

Hi Chris:

I make some experiments via calculations and channel blending...

A little bit of Surface blur in both cases in LAB mode.

Can you provide me a bigger versions of the files?

Maybe to my email...

I need to prove my theory to give you a better explanation...

Check my test at http: //tech.groups.yahoo.com/group/colortheory/files/Moire%20by%

Greetings from Costa Rica

Carlos Garro

___________________________________________________________________________

I don't really understand LEDs but your suggestion 2) sounds helpful. Another thing to try is a time exposure - say 10 seconds or more. May or may not help - it's just a guess.

Maris V. Lidaka Sr.

___________________________________________________________________________

Thank you for everyone's input. Here are answers to members' questions.

Thousands. I work for this client frequently, and at locations all around the country. Not only do they make billboards, they make LED "message centers" which are installed into constructed signs. As an example, think of a Las Vegas sign, with a backlit plastic cowboy wrapped in colorful neon, with a large 10'x20' LED display (somewhere on the construction, usually up 20' high or much more) with a relevant message displayed (e.g., Gamble Now!, Dancing Girls!, etc.). These things are fun to photograph, unless there's a nasty moiré pattern somewhere on the client's product.

This moiré problem is something that has recently come up as the pitch between the LEDs has decreased. The market trend for these displays is toward clean, clear video, and with that trend comes smaller and smaller pitches. In the examples I posted, the LED pitch is 19mm. This same company is currently developing tighter pitches in the realm of 13mm. It's no surprise that the target for these products is HD quality. They already use profiled color output (sRGB, NTSC or PAL profiles) and are able to get surprisingly accurate color.

No, but it's on my to-do list.

This looks very intriguing. Thanks for the link.

Heh. No kidding. ;-)

Yes, the moiré is embedded in the raw file, loud & proud. I'm using Canon 1Ds2 and even with its AA filter a moiré is produced. As photographed, the display is simply too large and dominant in the scene (it *is* the subject of the photo, after all).

Something else to point out. I never photograph these displays straight on (especially if they're mounted 50' high), so from an angle the frequency of the LEDs changes. The close LEDs are spread farther apart, the distant ones are tighter. This causes a change in their frequency which produces multiple instances of moiré within the same subject. I've added another example ("Mostly Blue") which shows the moiré produced with a wide angle lens.

They are in the "Photos" directory of this group:

http: //tech.ph.groups.yahoo.com/group/colortheory/photos

It seems to be a little glitchy at the moment, though. Try refreshing the page -- that works for me.

Again, thanks to everyone's contribution. I know this isn't solved yet, as far as digital capture goes. I will post any solutions that come my way.

Cheers ~

Chris Brown

___________________________________________________________________________

If anyone wants to take a crack at fixing a higher rez file, I've posted one at

http://www.chrisbrownphoto.com/recent

The file is "SurfSand.zip" and shows the moiré through text, smooth color and textured color.

BTW, this is not an attempt to get others to do this work for me. I've been hacking at this problem for a week without a promising solution, so decided yesterday to ask the professionals. The issue of moiré in digital capture is much more of a problem than it ever was with film, scans & halftone screens. IMO, it's one of the few major drawbacks to the technology. Perhaps it can be solved here!

Cheers ~

Chris Brown

___________________________________________________________________________

Chris,

Which converter are you using on your RAW images?

The reason I ask is that with ACR 3.6 (and the newer version of ACR present in the CS3 beta), some images display Moiré. Converting these same RAW files with Canon DPP, there's no Moiré at all (I'm using an EOS1D MII). Maybe a long shot, but worth a try.

I've witnessed this personally on fine fabrics (nothing like the LED displays your shooting, but you never know).

I first learned about this from a teacher in evening classes photography.

Regards -- Eric

___________________________________________________________________________

Chris,

I downloaded the file and had a look. I'd like to take a crack at fixing this, but want to get a shot at the cleanest data possible. Since raw converters interpolate differently, would it be possible to post the raw file on your site? If this is too large, then would it be possible for you to process the raw file with no sharpening and crop out a section at full resolution? Since the background isn't an issue, 10MB worth of the sign would be sufficient as long as it is a crop that does not introduce interpolation.

When photographing subjects with fine repeating patterns (imagine the speaker grill of many LCD/Plasma TV's), I've found that rezzing the file down sometimes introduces moire patterns hence option 2 in the above paragraph.

Thanks,

Darren Bernaerdt

___________________________________________________________________________

I took a look at your hi res file. IMO, you're going to need to do the following to get this to work:

1. Remove the pattern completely.

This is necessary because the pattern changes depending on the angle of the shoot and the LED distance and a bunch of other factors. This may seem impossible but about 2 minutes of using Neat Image removed almost all of it.

2. Add a pattern you created that doesn't moiré.

This will take some time, but once it's done it's done for all of your images. My coworker, Brandon, got a pretty good pattern that he set to Overlay.

I'd say those two things got us 90% there, and it only took 5-10 minutes.

--

J Walton

___________________________________________________________________________

Here's more info regarding the latest questions:

Capture One Pro (latest version)

Done. The file is named "Billboard_Raw.zip"

The first image I posted is at max resolution, without any USM applied in the processing software (Capture One). Nothing was done in Photoshop except save it as a PSD file.

I agree, reducing the image size can "hide" the annoying pattern but in this case the client is interested in reproducing large images for its product announcement materials. By "large", I mean 12"x18" @ 300 dpi.

Thanks again for everyone's gusto. This issue has beat me up a bit over the last week.

~ Chris Brown

___________________________________________________________________________

On Mar 14, 2007, at 11:00 AM, J Walton arranged some pixels so they looked like this:

What do you mean by this? Can we see an example?

thanks,

_andrew webb

___________________________________________________________________________

Hello again, it's been a while since I've posted but I have been lurking in the shadows. I do learn SO much from everyone, thanks!

This may or may not be a possibility; compositing the location shot with the file that was used to create the sign. I have been following this thread and it just seems that, aside from going to film, a hi- res digital capture will result in the moiré pattern consistently.

Not sure what type of file is used to create the illuminated sign but have been curious as a couple sign have been appearing around western NC. It would be a very simple PS task to composite the two.

Just a thought.

Cheers!

Tom

___________________________________________________________________________

On 14 Mar 2007 19:28:49 -0700, Andrew S. Webb wrote:

Sure.

Make a new document, and run the Texture > Patchwork filter. Use a steep shadow curve to emphasize the vertical lines. Or open one of the many Postscript patterns in the Photoshop Presets folder - plaid, diamonds-cubes, herringbone, mayan bricks, waffle, rough diamonds - almost anything small could potentially make a good texture. Fill a larger document with the smaller pattern.

Or (which is what I'd do) create a texture yourself that you can repeat and that will roughly look like LEDs. Check out one of those canned Photoshop tutorials that teach you how to make television lines. You just want something that you can repeat and distort to fit the billboard and set to Overlay or Hard Light or something.

The key to this job is to remove the pattern & moiré altogether (I recommend Neat Image or the original digital image for that matter) and put a "good" pattern on top. I think once you found a pattern that worked well at a certain size you could process each file in 15-20 minutes.

The smoothed-out image + noise + a clean pattern + a halftone screen should look just fine.

If that's not acceptable because the client insists on the actual LEDs making up the pattern then the only thing I know to recommend is film. But what the client doesn't know won't hurt you if you do a good enough job on the texture and don't tell them.

--

J Walton

___________________________________________________________________________

"I didn't realise this is scanned film. Paul's post brings something back to mind - can it not be eliminated by selecting a lower scan resolution?"

This is not scanned film, it's digital capture from a Canon 1Ds2.

~ CB

___________________________________________________________________________

Is the large image you posted a 1:1 crop from a full-sized image, or was it downsized? I'll assume the former, since it's not 4:3 aspect ratio.

Some Canon cameras, at least, are known for having pretty heavy anti-alias filters, and people complain about their softness. I wonder if the 1DsMkII is different in that regard. I've never seen aliasing on my lesser Canons. Of course, shooting an array of LEDs, where the spatial frequency of the LED array is getting close to the spatial frequency of the sensor array, is absolutely the worst case.

If you back way up until the camera can't see the individual LEDs any more, then obviously there will be nothing to alias, but that's probably not a satisfactory solution. If you get closer, so that the sign fills the entire frame, then the lower spatial frequency of the image might yield less aliasing, but that may also be impossible.

An anti-alias filter is actually a diffuser, which turns a point of light into a fuzzy blob. This is the only mechanism that actually eliminates high spatial frequencies, and prevents moire. However, you might try simply defocusing a hair, not enough to make the LEDs disappear in a complete blur though. While defocusing isn't mathematically the same as diffusing, it might help.

Ciao, Paul D. DeRocco

___________________________________________________________________________

Could this moire problem actually be caused by the LED cluster itself? If there is any sort of polarized glass in front of the cluster, to protect the electronics, the interference between the polarize frequency and the LED frequency may be causing the problem. Does this pattern show up on the source film?

Chris Corich

___________________________________________________________________________

You will have problems with moire whan you go to halftone unless your feature size (i.e. the size of each LED in your image) is considerably larger than the screen pich. But colour screens for offset printing are quite fine anyway, so if you want people to see the LED pattern you'll have to make sure each individual LED is big enough.

Andrew.

___________________________________________________________________________

Time exposure won't help. Scanning film will probably change the frequency relationships.

The only way to abate it without blurring is to change the frequency relationship between the LEDs and the pixels to one that generates less visible interference. Door #2 is on the right track.

Randy Hoffner

___________________________________________________________________________

The OP did not say he was shooting film so the assumption, these days, is that he shot digital. I suggested that shooting film might solve the problem. The OP said that he would be doing this sort of job on a regular basis...so the problem would keep occurring making it worthwhile to find a good solution.

My instincts would be to scan at a higher resolution to prevent moiré but only after trying to find an angle at which to scan the film where the moiré did not occur.

Since moiré happens when multiple regular pattern interfere with each other I have to wonder if there is any way to predict if a moiré would happen when the image is halftoned for printing.

___________________________________________________________________________

Thanks again for everyone's interest in this problem.

Several posts suggested mocking up the content of the sign either by using the file of the original art/ad. Unfortunately, this is not a viable option for two reasons. In most cases because we will not have access to those art/ad files or be able to get them in a timely manner. Secondly, we don't want to call up the end user of the sign and ask for the ad file to help in mocking up the LED display. That would only send the message that the displays are crap and that all the marketing materials from the LED display company use faked & retouched product shots.

As I mentioned earlier, the fact that the LED matrix *is* the product prohibits us from manipulating the scene too much (blurring, blending, etc.) if it deteriorates what's displayed on the product.

Some of you may still be trying post-production solutions on the files I posted, but currently it looks like the solution is one of the following:

1) Shoot film & scan it. This would need to be tested for moiré patterns in the proofing stage.

2) Use an "anti-moiré" filter (I have not tested this yet but have contacted Caprock).

3) Shoot digitally from a distance that reduces the LED matrix to a size such that there is no moiré. This limits the reproduction size of the photo but may be a viable solution in some cases.

For those of you lurking but want to hack at this problem, sample files are here:

http://www.chrisbrownphoto.com/recent/

Billboard_Raw.zip is a raw file (Canon .cr2)

SandSurf.zip is a Photoshop file of the same raw file, simply cropped. No USM or anything else has been done to it.

Cheers ~

Chris Brown

___________________________________________________________________________

Chris,

I tried a number of different blends on the L channel to try and knock down the moire. (Blurring the color channels was the easy part.) I found it is possible to virtually eliminate it, however the amount of work involved with making masks of each color would be prohibitive in a production oriented workflow.

I ran your RAW file through a variety of software to see how that might change it. It seems that C1 gives the strongest rendered moire. RAW Developer with the "Early Stage" and "ChromaLogic" noise reduction cranked up made the moire much less objectionable. Still there, though.

Since taking a digital file of the display, adding a pattern and dropping it into your photo isn't an option, I would lean to techniques that can be used at the time of photography. It sounds like you just have too many of these to shoot to get bogged down in Photoshop.

As you mentioned, film or the Caprock filters might be a successful route to go. Since you are not getting any moire on the right side of the display, that seems to be a workable camera to subject distance to avoid the moire. Maybe shooting it more head on at that distance would work. I also found using a less than stellar lens would degrade the image enough to avoid moire in some situations. One last thought might be to use a DSLR with smaller photosites like a Canon 30D, Nikon D2x or Nikon D200.

One last comment, depending on the technology being used for your prints, you may experience moire at the printing stage even if it is not present at 100% on your monitor. I was printing some portraits on an Epson 4800 that exhibited moire. There wasn't any present in the file, however the resolution of the images versus the screening/resolution of the printer interacted to produce moire. Dropping the resolution on the image changed the pattern frequency in the subject's jacket and that cured the moire.

Isn't digital great? ;-)

Good luck and let us know how this works out.

Darren Bernaerdt

___________________________________________________________________________

Having thought about this some more, I think the problem would be lessened by shooting film and scanning. I think you're getting into trouble because the camera uses a Bayer pattern sensor (unlike a scanner), LEDs are monochromatic, and the anti-alias filter in the camera is really tuned for white light affecting all three color sensors.

Ciao, Paul D. DeRocco

___________________________________________________________________________

I agree with those who suggested that there are ways of shooting this that would substantially reduce the moire if not eliminate it altogether. Even as it stands, though, the moire is neither particularly easy nor particularly hard to remove.

The annoying problem is that it can't be removed as a whole because it has different characteristics where it crosses different background colors. This mandates one pass for the blue background areas and one for the red, etc. As against that, the pattern is so well defined as a matter of color that a lot of blends become possible.

You'd need a selection (or Blend If, if one suggests itself) of the affected color area. Then, as Darren suggests, convert to LAB, where the moire can be easily obliterated in the A and B channels, which are very forgiving. The problem of getting it out of the L channel is addressed by blending with a modified version of the A or B.

I took the moire out of the blue areas by the following:

1) convert to LAB

2) make duplicate layer

3) apply Surface Blur to top layer strongly enough to obliterate the *color* moire in the blue areas. The moire in the L is reduced but not eliminated; other areas of the image look bad.

4) Use Blend If to restrict the move to the blue areas of the image by choosing This Layer and excluding everything that is not negative in the B channel.

5) Going to the bottom (original layer) compare the three channels to see whether the A or the B has a more similar moire to the one found in the L. In this case, it's the B.

6) Pull a separate copy of the original B. To this copy, apply a lightening curve to make the blue background (the area unaffected by moire) as close to 50% as possible.

7) Return to the top layer of the other document and activate the L. Apply Image, source the separate B channel, mode Hard Light. If it aggravates the moire, check invert, which will lessen it. Lower the opacity until the moire is minimized (otherwise it will put in a counter-moire).

8) If necessary, Surface Blur the L again after this blend.

When I did this to the high-res file the moire was basically gone in about two minutes of work.

Now, as for capture:

Moire is minimized when the angle of capture is 30 (or 60) degrees away from the angle in the source. Probably the most foolproof way is to shoot film, then scan it at a 30 degree angle (plus or minus any variation in the original photograph--the billboard isn't exactly horizontal in the original file.

If shooting digitally, you would definitely get a better result by angling the camera by 30 degrees as well.

Dan Margulis

___________________________________________________________________________

Not so sure about this. CCD scanners do use Bayer pattern imagers, generally. The Bayer pattern will, I think, cause a different moiré pattern for different colors of LED: for example, green vs red, as there are twice as many green pixels as red.

Drum scanners I don't think use Bayer pattern sensors, but the image is still chopped up into pixels when it is sampled and digitized, so I think moiré would still be a problem.

And I believe that an anti-alias filter is probably just a piece of ground glass or something that filters out the higher frequencies optically.

Randy Hoffner

___________________________________________________________________________

From: Hoffner, Randall N

What CCD scanner uses a Bayer pattern? My old LS-2000 certainly doesn't. I'm pretty sure that the Epson flatbed "MatrixCCD" technology uses six lines of three colors, two lines per color.

Depends upon the spatial frequencies involved. A scanner may have so much resolution that, even with no anti-alias filter, it'll accurately represent the image with no aliases. For instance, the V750 runs 6400dpi, which on a 35mm slide is about 60Mp.

Exactly.

--

Ciao, Paul D. DeRocco

___________________________________________________________________________

Thanks. Makes sense now.

Cheers,

_andrew webb

___________________________________________________________________________

If it were me and I still had my view camera I would do it this way. Shoot 4x5 or bigger. Use swings and tilts to fix perspective. I think I would shoot transparencies (slide film) as I only need one of each sign so individual processing is not a problem and it can be done with minimal equipment. Use a digital camera to check the exposure. Develop for flexibility rather than accuracy.

This would give you so much information that you can test lots of different scanning techniques.

Gene Palmiter

___________________________________________________________________________

This is a very fascinating problem, that I unfortunately have found no solution to... I grabbed the RAW file and have had no success so far in removing the moire without destroying the sharpness or detail in the process. I might be able to do better by throwing a lot more time at it, but as someone else mentioned, in a production workflow you can't spend hours on each image, you need to find a workflow that works, quickly, for the thousands of images you're going to have to process.

I would definitely try shooting to film; my concern is that since the LED grid is still present the moire is likely to show up when the files are halftoned to make the film (or press plates, in a CTP environment), much further down the line than you would want. I would definitely test this process all the way to film or plate output.

I see nothing wrong with asking for the original layout files for these billboard images, and I don't think it's going to send any wrong message to those folks once you explain *why* you need them. Computer monitor and TV advertising use simulated images all the time (and most often will include a tagline to that effect somewhere in the ad layout) for very similar reasons, namely that there is nothing wrong with the displays themselves, but getting a photograph that accurately shows the quality of the display without artifacts is extremely difficult.

That been said, I don't know if obtaining the source files for all of these will be a timely endeavor, either.

I look forward to reading more suggestions on this... good luck!

BRIAN PYLANT

Manager, Electronic Prepress

:::::::::::::::::::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::

Disc Makers

7905 North Route 130, Pennsauken, NJ 08110

___________________________________________________________________________

Following this thread has brought me to this conclusion of what I would try if faced with this dilemma

I would shoot multiple frames of the sign at different zoom ratios. It seems the moiré pattern occurs at certain frequencies between the led pattern and the Bayer pattern.

Multiple zoom shots should produce moiré in different areas of the sign.

I would then stack these different shots as separate layers and then resize the various layers to match the content of the basic overall shot. I would use layer masks to isolate and/or block moiré/non moiré portions of the individual layers.

Regards

Ray

Raymond St. Arnaud

Audio Visual Services

Camosun College

___________________________________________________________________________

Been thinking what I'd do too. I keep coming back to the surface of the billboard and lights creating this effect. I'd try a polarizing filter. And assuming it will be hard to see this moiré while taking the photo I'd manually adjust (turning the outside glass) it during a couple of captures.

Lee Clawson

2/\V/\7 Studio

___________________________________________________________________________

On Mar 16, 2007, at 9:02 AM, Raymond St. Arnaud wrote:

I agree. The other obvious solution would be to shoot with a camera with a 4-shot sensor or use a scanning back like the BetterLight. The higher resolution would also reduce the aliasing (moiré).

--Rich Wagner

___________________________________________________________________________

From: "Raymond St. Arnaud"

I like this idea. Since the moiré is in the form of dark bands, it may be sufficient to combine the layers in lighten mode.

Mike Russell

www.curvemeister.com/forum/

___________________________________________________________________________

Again, thanks to everyone for their suggestions. Here is some current information on this situation.

In order to produce something quickly for the ad agency working with this client (the LED display manufacturer), they sent me the actual files used to for the display, which were overlaid and blended to retain the "dots" of the LEDs and hide any distracting moiré pattern. This was a solution that one group member suggested, and had already been discussed between the agency, client and I as a quick, stopgap solution.

The agency loved this solution because the colors are more vibrant (i.e., fake) than what the LED display produces. The client is a purist and doesn't want this to be the final solution. In addition, this solution can't be used for billboards for which the art files are unavailable or inaccessible.

In the next few weeks, I'll be testing film+scan, which the client wants me to try simply for the sake of comparative evaluation. I have no doubt that any moiré will be eliminated in the file, but may rear its ugly head if/when the file is resized and printed in halftone.

Here are responses to others' suggestions:

The moiré is absolutely generated by the interfering frequencies of the LED matrix and camera sensor's matrix. We (the client & I) have known this and have tried altering camera orientation and angle to manipulate the frequency of the LED matrix without success. This "zoom" solution crossed my mind, and would not require a change in camera position to alter the frequency of imaged LEDs. The only drawback is that for each billboard there will be many ads displayed that will be photographed, so the total number of images would increase by a factor of 5 or 6 when "bracketing" for moiré.

The file "SandSurf.zip" was taken with a B+W circular polarizer, but it was used to enhance the sky and darken the black surface of the LED panels, not interfere with the LED frequency. I doubt a polarizer would have any effect because it does not change the frequency of the LED matrix.

I doubt this solution would work, but haven't tried it. A moiré pattern is generated by interfering frequencies of the pixels of the LED matrix and camera sensor matrix. A 4-shot back still has a sensor matrix, with a specific CCD pitch (frequency). As long as a sharply defined matrix exists in both the scene and the imaging sensor, there exists the possibility of interfering frequencies. Imagine taking two screen doors and laying one on top of another. As the top one is rotated, the moiré pattern will change and transform shapes, but will not be eliminated until the screens' grid is exactly aligned. If the top screen is changed to a finer mesh or coarser mesh, the moiré pattern will still be there, just in different patterns due to the different frequency of the mesh.

One I have hacked at several more solutions, including film+scan, I will post to the group. In the meantime, if anyone wants to post their solution to the groups "Photos" directory, please do so.

Cheers ~

Chris Brown

___________________________________________________________________________

One of the big claims to fame of multi-shot backs and scanning backs is the absence of moiré. Try a Google search for (multi-shot and moire) or (BetterLight and moire).

With scanning backs, the pixel frequency is far higher than your Canon, and to effect the image is over-sampled, reducing or eliminating the alias that is called moire. There is also not interpolation of lower frequency sensor data to generate the Bayer pattern. The scan back does not produce the same moiré or color aliasing that is typical of single-shot digital cameras. A Super8K will produce files with 8,000 x 12,000 resolution.

BetterLight just released a 416 Megapixel model this month. <http: //www.pdn-pix.com/pdn/prodtech/news/article_display.jsp?vnu_content_id= 1003557675>

"Better Light scanning backs are widely acknowledged for capturing image quality superior to film or fixed-array digital cameras. Used on a 4x5 view camera or any device accepting 4"x5" film holders, the capture area is 72mm x 96mm (120mm diagonal), creating a huge pixel population with the scan movement of its tri-linear sensor. Each pixel receives pure red, green and blue data so there is no interpolation of color data, or typical digital artifacts, or moire pattern issues to deal with in post-production.

The new Super10K-HS answers the call of museums and similar institutions for larger native file sizes. It has the ability to create a direct digital image at a native resolution of 10,200 x 13,600 pixels. This high resolution allows scanning of originals up to 34 x 45 inches in size at 300 pixels per inch without stitching or interpolation, and generates a 794 MB 48-bit RGB file size."

You might see if a rep would demo one for you.

--Rich Wagner

___________________________________________________________________________

Ahem, don't forget the Foveon sensor. It doesn't have the raw megapixel power of the multi-shot and scanning backs but it'll produce a nice moiré free image in my experience.

Regards,

Terry " the OTHER guy who bought a Sigma SD-9" Wyse

:-)

_____________________________

WyseConsul

704.843.0858

http://www.wyseconsul.com

http://www.colormanagementgroup.com

___________________________________________________________________________

I have rented a Better Light for museum work and was impressed with the results, but not with the amount of light needed (or the fixtures, for that matter).

Unfortunately, the content of these billboards is displayed for 15 to

30 seconds in real world use (the messages I captured were displayed for 5 seconds each). My experience with scanning backs was with a 2- to 4-minute exposure. With smaller signs, the content is usually displayed for less than 5 seconds per message, to get the most messages to the audience. As the pixel pitch gets smaller and the image is more "continuous tone", the message content is more often a video or 3-D motion graphic which is in itself a challenge for still photography.

Believe me, if I felt a scanning back or multi-shot back would work, I'd try it.

Side Bar:

It's almost tragic how much work goes into digital capture workflow to avoid its pitfalls. With film, this problem would only be an issueat the scanning stage, and a temporary one at that.

Cheers ~

Chris Brown

___________________________________________________________________________

On Mar 18, 2007, at 2:33 PM, Chris Brown wrote:

Well, that kills the scanning back idea.

Absolutely. Now this is getting to sound impossible on digital capture, and still tricky on film.

I believe you! It looks like it's time to go back to film for this project.

For this problem, that's certainly true. But I still have a freezer full of film, and no intention of going back, unless I'm forced to for a particular project.

Overall, the advantages outweigh the disadvantages, but it sure has changed the way we work.

Good luck! This is a very challenging project.

--Rich Wagner

___________________________________________________________________________

I admit ignorance here, never having worked with a device of this nature, but is this not a setting you can change just for the duration of the shoot?

BRIAN PYLANT

Manager, Electronic Prepress

:::::::::::::::::::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::

Disc Makers

7905 North Route 130, Pennsauken, NJ 08110

Toll free: 1-800-468-9353 ext. 5539

___________________________________________________________________________

There's another factor here, too, that has not been mentioned. I strongly suspect that these leds are driven with multiplexed signals: even when they are "on", they are not continuously "on", but are flashed on and off at some high frequency. I'd almost be willing to bet on it. As Chris is well aware, moiré is caused by interference between two or more frequencies. The severity depends on the relationships between the various frequencies. As he is also well aware, it is pretty hard to avoid a clash of frequencies in this kind of situation, somewhere along the line.

Television engineers say, "Welcome to my world".

Randy Hoffner

___________________________________________________________________________

In the case of my examples, that sign is an installed demo unit behind their factory where I could exert control if needed. However, in real world use, the content of a given sign is displayed for a duration that is tuned to the traffic flow. Slower traffic gets shorter message times (to show more ads), faster traffic (e.g., interstate highways) get longer display times. In some cases, the times of duration for an individual display are dictated by local laws. In addition, the host computer is typically in another state, and is controlled by the billboard company (e.g., Lamar, Comcast, etc.) via cell phone or DSL connection.

In the last year or two, refresh rates for LED signs have topped 500 Hz. Before that, most signs were refreshed at only 1/30th of a second due to legacy electronics.

However, the LEDs rarely run at 100% brightness (due to color management) so it takes several refresh cycles to get adequate brightness for a decent photo.

Cheers ~

Chris Brown

___________________________________________________________________________

Chris

Sorry this is late

Boy you have lots of good advise. Now for something Expensive, (client pays for), and something funky.

You might try renting a digital back with a four pop exposure mode as opposed to One. This helps the Bayer matrix stuff. These Four pop camera backs are always better for moire patters, this I know from shooting clothing. And you might as well go for the 39 or more megapixel backs like Sinar and PhaseOne. Prey for no wind, ..... or let the movement in the trees be "Art".

On a funky note: To get more megapixels (thus changing the ratio size of LED to C-Mos pixel), do a panorama of the Billboards and stitch the image together :)

I have never had luck changing the camera rotation to eliminate moire. Yes, it looked good at an angle on the monitor but when re-straighten the moiré was back. To be fair I didn't measure 30

degrees exactly, I tried a number of off the cuff angles.

David B. Moore

Photographer

http://www.azphotobook.com/davidmoore/

C 480-751-7689

5845 E Onyx Ave

Scottsdale AZ 85253