M8 at f8 on a California Freeway and Moire at 60mph!

[Asher Kelman]

Of course the Leica M8 can be hand hwld at 1/8 sec to get sharp images at a wedding or standing on a street corner.

I also like to test cameras in the way I use them. I'm unorthodox. I am known to shoot zone focussed while walking, not even stopping, or out the car window.

So I was pleased with the sharp wonderful pictures the M8 captured at f8 with the 28mm f2.0 lens.

What surprised me was Moire!

If you drag off the file to your screen and increase or decrease it a little the moire will become more pronounced.

It occurred on many shots!


Asher

 

[Ray West]

Asher,

I think the Moire pattern you see is a function of the image and the vdu screen dot resolution. It would be interesting if you print it, see what size you have to print it at to get the same interference patterns.

Best wishes,

Ray

 

[Asher Kelman]

Asher,

I think the Moire pattern you see is a function of the image and the vdu screen dot resolution. It would be interesting if you print it, see what size you have to print it at to get the same interference patterns.

Best wishes,

Ray

Just print it and you'll see the Moire! I justvprinted that file at 8x10". The moire is obvious. One can adjust by going up and down in mag factor but it is a clear problem!

Asher

 

[Ray West]

Hi Asher,

As you know, Moire is the result of similar regular patterns being superimposed, Wikipedia has some info (and calculations) but plenty more out there. It is a compromise, if you make digital cameras, which have a regular sensor pattern, as to how you blur the image, before it hits the sensor. It will happen, sooner or later with all sharp digital camera systems. If it seems that the weaker filtering on the M8 allows moire to occur more frequently, and if Leica do something about it, then the image is likely to suffer in some other way. If you downsample, or upsample the image, choose a different print resolution, I think you can diminish the printed effect, in a similar way as zooming on the screen alters it. Anyway, I suspect selective treatment in pp will cure it.

At the moment, it looks as if the M8 is only reliably useful for natural landscapes. I would be interested to see a shot with the M8 showing moire, compared with the same shot from a 'lesser' camera.

Best wishes,

Ray

 

[Ray West]

Hi Asher,

If I zoom out on your image, then I get the moire as a result of the 'road lines' interfering with the vdu screen dot pattern. This pattern moves, depending on the image size. If I zoom out, I get the moire pattern within the image itself. I do not know if jpeg adds to this, but at high magnification, I can see that the fringe pattern is blurred, whereas the normal pattern is pixelated. At the normal size, the pattern is not that noticeable on my screen.

This puts another factor into the equation, the image conversion method. (I know your original would be raw, but even that is 'converted' to some extent.)

Best wishes,

Ray

 

[Asher Kelman]

Hi Ray,

All one is doing when you change magnification is altering the coincidence of cycles. Moire will peak, then gradually go down then peak again in cycles as you increase or decrease the magnification. One can shift the moire up and down the patterning of the actual image.

However, it is always there.

No need to obsess just print the file.

I even printed directly from the PSD file at 360 dpi without any manipulation and we get the same moire in the 8x10 print!

Try it for your self! It's the M8 file!

Asher

 

[StuartRae]

This puts another factor into the equation, the image conversion method.

Asher,

Assuming you did save the image in RAW format, have you tried different converters?

I know that RSP was often guilty of colour artifacts, but this could be cured to some extent by decreasing detail extraction. Alas RSP is no more and so doesn't support the M8.

Regards,

Stuart

 

[Ray West]

Hi Asher,

Maybe I wasn't clear in my last post. I am saying that my initial response, re the zoom etc. is connected with the screen resolution and the near parallel lines in the road surface. That sort of interference can usually be removed by printing, or viewing, at different resolutions. This gives the moving patterns, it is not related to the camera sensor (although if the sensor did not resolve the lines in the first place.... ). However, if you zoom in on your image, beyond the pixel level, you can see the moire due to either the camera sensor itself, or the jpeg conversion or even the raw conversion process. the 'rings' are clearly blurred (oxymoron ;-) .

I have an image here, http://www.openphotographyforums.com/forums/showthread.php?t=1251which has similar artifacts (penultimate posting). If my original is in raw, I will see if different convertors change it, as Stuart indicates.

Best wishes,

Ray

 

[StuartRae]

In his article about the Nyquist frequency and aliasing, Norman Koren suggests that aliasing can be reduced by stopping down to about f/16, at which point diffraction acts as an AA filter.

Regards,

Stuart

 

[Nathaniel Alpert]

There seems to be tremendous confusion among photographers about antialiasing filters and digital imaging. On the other hand, this phenomena is well understood in digital signal processing; all of our digital toys are desigend according to the Shannon sampling theorem. One way to think about this issue is to imagine that a "virtual image" is projected onto a plane at the level of the sensor. This image is "sampled" by the sensor at a finite number of locations. If the virtual image has no information above the Nyquist frequency of the sensor, one can "reconstruct" the virtual image exactly, assuming the image is noiseless. If the virtual image contains significant information in the frequencies above the Nyquist frequewncy of the detector, there will be alaising (Moire). It will be more or less objectionable, depending on the scene (e.g. fabrics) and the quality of the lens. The engineering solution is to interpose a filter that will cut off all information above the Nyquist frequency of the sensor, on the grounds that there is no free lunch. This works, but the filters are not perfect; they cannot cut everything off without some blurring below Nyquist. Without a proper antialiasing filter, the M8 will exhibit alaising. It is there all the time, not just when you image a fabric or hairs. Aliasing is a distortion that places high frequency information in the wrong place in the image. One can argue that this will not be objectionable in a landscape because it is spread throughout the image, but I think that a properly designed camera with a well-designed AA filter and a good lens should outperform the camera with an equally good lens and no AA filter. The problem in evaluating this assertion is that in the real world, it is difficult to remove the lens from the evaluation. Measuring the same camera and lens with and without AA filters would be the best experiment.

' Hope this helps.

 

[James Roberts]

{snipped} The problem in evaluating this assertion is that in the real world, it is difficult to remove the lens from the evaluation. Measuring the same camera and lens with and without AA filters would be the best experiment.
Hope this helps.

Nathaniel--it's probably an experiment that Leica did already, don't you think? Or--purely conjecture here--could there have been some other optical reason for the very weak AA filtering (I'm not entirely sure about "lack") in the DMR and M8?

All I know is practically with the DMR I don't run into artifacts that mar prints any more than I would with my 1ds2 (yes, in C1 from time to time I'd see moire there too--especially in fabrics).

Asher--I haven't had time to print your shot; I think the suggestion for different converter is a good one, though.

 

[Ray West]

Hi Asher & Stuart,

fwiw, my original image was a crw file, my version of canon dpp does not work with it. However, in Silkypix, Light zone, cr2, & even Irfan view, all set to 300% magnification, there is only a slight amount of moire visible - there are worse things going on. However, it is clearly seen that it is about the mid image area, where the sensor gives up resolving the ridges in the walkway. This is exactly as I would expect it to be, being so wise after the event. There are differences in the colours/noise that each raw viewer shows, but in this particular moire test, they all show more or less the same artifacts.

Now, my jpeg version, 'freezes' the image at whatever resolution is chosen, and picks up the moire pattern, in a similar way it would be shown on screen at that resolution. If you zoom in, you see that the pattern stabalises, showing that captured moire, which is far more moire than in the raw image, what you see in the jpeg as moire has little connection with the sensor. If you zoom out, you see the moire effect, as the thin straight lines interact with the fixed pixels of your monitor. I think the same may apply to Asher's image too, maybe we could have a peek at the raw image.

Best wishes,

Ray

 

[Nathaniel Alpert]

Issues about aliasing are covered in digital imaging 101. It is hard to believe that Leica did not know exactly what they were doing. I don't think it is useful to speculate on their rationale.
On the other hand, I have read that the physical space between lens and sensor in the M8 makes it difficult to include an AA filter (and IR as well) but I don't know that for a fact. Life is full of compromises.

 

[Asher Kelman]

I am a great advocate of the M8. I can use it where the 5D with lenses is more noticable. The bag with lenses weighs nothing and the files are wonderfully robust and the pictures great!


BTW, it prints with moire at 300 dpi from a PSD file with no uprezzing of the file size!

I'll send anyone the file for processing or curing to the extent possiblle!


Cheers!

Asher

 

[Don Lashier]

Asher, have you tried C1's moire filter? It usually does a pretty good job.

- DL

 

[Asher Kelman]

Remember, this is an excercise and I do not see it as a criticism of the camera, since I have not as yet found that it hurt anything I seriously wanted to do. Yet I'd like to know about it so I can work around these phenomena.

Here's the file for download!

Please pm me that you have downloaded it OK.

Thanks so much for having a go!

Asher

 

[Ray West]

Hi Asher,

The only Moire I see, is in your lane, just behind the vehicle in front of yours, at the edges, where the grooves in the road surface become blurred. It seems to be red and green fringing. I have only loaded the file into Lightzone. I have had a play, but not good enough to give a definitive answer, since I am not an expert at the capabilities of lightzone. I think the easy solution, would be to select the region, being generous, and using a broad feather, and then blur it. The road beyond is blurred, i.e. the camera sensor can not resolve the grooves, you're simply bringing that area forward a few yards. I guess other solutions would be to clone the smooth road forward, but I couldn't get LZ to clone sensibly. It may be, that some of the chromatic aberation fliters in more advanced editting software could do it. As it is, it is not noticeable at 'actual pixel size' on the screen. It is exactly tthe same characteristics as in my image, mentioned previously, in that it is not the same in raw as in the jpeg, and the same sort of area gives the fringing.

I think, compared to the other qualities in the images that the M8 seems to produce, I'd prefer to keep the colour range and detail, rather than sacrifice it for reducing the odd bit of moire fringing. There is a fair amount of detail in there.

Best wishes,

Ray

 

[Nathaniel Alpert]

Remember, this is an excercise and I do not see it as a criticism of the camera, since I have not as yet found that it hurt anything I seriously wanted to do. Yet I'd like to know about it so I can work around these phenomena.

Here's the file for download!

Please pm me that you have downloaded it OK.

Thanks so much for having a go!

Asher

The fact that the M8 does not have an effective (any?) AA filter means that it will always exhibit aliasing. Aliasing (Moire) cannot be removed by any filter without additional information about the scene. What this means is that although the images from the M8 may be better in some way than those from other cameras, because of aliasing they are not as good as they could be. Ordinary photgraphers cannot evaluate the sensor without also considering the lens as part of the system. Lieca lenses are legendary, they may be contributing much of what people like about the M8 pictures, due to their perfromance below the Nyquist frequency of the sensor. On the other hand, very sharp lenses pass more high frequency information to the sensor and thus there may be more trouble with aliasing. It is important to understand that there will always be aliasing in every image, the only question is whether its presence will outweigh the other virtues of the M8.

 

[James Roberts]

I have seen no trouble to date with aliasing and the M8. Quite the opposite; so far, it's doing as well as the 1ds2 and perhaps even better with detail.

 

[Asher Kelman]

Jamie,

Have you looked at this DNG file? Here's the file for download! It would be interesting to see whether or not you see the moire as a file issue or an interaction with the screen or printer to get the moire.

It certainly is visible and always prints out from a PSD file.

Asher

 

[Will_Perlis]

There's some component that's related to the screen and magnification. Here, in PS with a 1280x1024 LCD at 33% it's very apparent, at 50% it almost vanishes and it comes back strongly at 66%. I'll be curious to see what happens on the 1600x1200 LCD.

 

[Ray West]

Hi Asher,

The areas where you have fringing due to the camera sensor and the road surface are within patches which I've shaded green, below. Check it on your raw convertor, see if its the same as what I see. Keep zooming in, until the pattern stabalises. (probably red and green swirls, roughly within the area shown)

All other artifacts, outside of those patches, when printing or viewing on screen are due to interference patterns between the image and your print/view device. Unless you zoom in on the image, as I keep mentioning, you could be talking about moire patterns other than those caused by the actual object and camera sensor. I do not think it is a significant problem, unless you are upsizing the image, since it can be cloned out. I do think you may have a problem in finding a suitable print resolution to overcome the natural interferance between the lines in the road, and your print head. I wonder if a dye sublimation print would be different.


Best wishes,

Ray

 

[James Roberts]

I don't think this is moire...

Asher--I've looked at the file enough to develop it out of C1 and print a quick test (ImagePrint, 4000, 8 x 12 @ 300ppi).

I'm not seeing a "moire" pattern on the prints so far.

I also don't see moire, I don't think, on my monitor (1280 * 1024 EIZO) looking at the developed TIFF @ 100% in PS CS2.

I see the moire effect of the screen and road at almost any other magnification. But PS does funny stuff at more or less than 100% magnification.

What I *do* see on the monitor at 100% is some very subtle grey-green-ish artifacts right at "focus extinction" (where those parallel lines stop becoming lines). This could be some kind of aliasing, but it's very subtle. This is probably what Ray is talking about...

And I'm not sure I would have noticed it at all if I hadn't been looking for it--the artifacts are not printing exactly the same way they're displaying.

Of course, I'm also not sure there isn't some weird visible artifact in the shot actually there on the road--that diagonal white line "scratch" on the concrete is actually pretty astounding detail. I guess it could be something on the windsheild, too... you did shoot this through a windsheild, right Asher?

I tried the C1 moire plug-in and the on-screen effect actually got worse after the plugin completed, which is also strange.

 

[Asher Kelman]

What I *do* see on the monitor at 100% is some very subtle grey-green-ish artifacts right at "focus extinction" (where those parallel lines stop becoming lines). This could be some kind of aliasing, but it's very subtle. This is probably what Ray is talking about...

Of course, I'm also not sure there isn't some weird visible artifact in the shot actually there on the road--that diagonal white line "scratch" on the concrete is actually pretty astounding detail. I guess it could be something on the windsheild, too... you did shoot this through a windsheild, right Asher?

No I shot out of the right passenger window! Outside, totally outside!

I tried the C1 moire plug-in and the on-screen effect actually got worse after the plugin completed, which is also strange.

James, this, I believe is real.

BTW, Will Perlis first used the camera in Samy's Camera in Los Angleles. He immediately had severe moire in the first pictures. I have had a different experience until now.

All my pictures are beautiful. The artifact is real but it doesn't take away my admiration for the M8. Just look at the New York Central Park pictures of Steve Teitelbaum.

I'll do some more work on this.

The moire prints out easily from CS2. Try it at 300 dpi. File prepared in Adobe RAW.

Still puzzled but a fan!

Asher

 

[Bart_van_der_Wolf]

Have you looked at this DNG file? Here's the file for download! It would be interesting to see whether or not you see the moire as a file issue or an interaction with the screen or printer to get the moire.

Asher, it is camera induced moire, no doubt whatsoever.

Since the red and blue filtered sensels sample the image projected on it more sparsely than green-filtered ones, there will always be a difference (whether AA-filtered or not) in the onset of moire visibility as one approaches the so-called Nyquist frequency. This will result in false color artifacts, the amplitude of which also depends on the Raw converter used.

An AA-filtered sensor, probably impossible to implement with such close proximity between lens exit pupil and sensor as in the M8, will only reduce the aberrations. Total eliminaton would result in significant loss of resolution, a trade-off not chosen by manufacturers of traditional Bayer CFA designs. A more gentle roll-off is common, although there are differences between designs.

The lack of an AA-filter will always result in more Aliasing artifacts, but they may not always be disturbingly visible. An AA-filter will always reduce the amplitude of Aliasing artifacts, but also reduce the MTF of the system's limiting resolution (although some of that reduction can be restored in Postprocessing).

Bart

 

[Bart_van_der_Wolf]

There's some component that's related to the screen and magnification. Here, in PS with a 1280x1024 LCD at 33% it's very apparent, at 50% it almost vanishes and it comes back strongly at 66%. I'll be curious to see what happens on the 1600x1200 LCD.

The effect you see is caused by (quick but inaccurate) down-sampling without proper anti-aliasing precautions. That 'zoom-effect' is most visible at reduction factors other than 1/2 x

Bart

 

[Ray West]

Hi James,

Thanks for confirming what I see on the original image. The rest of what is talked about here, is nothing to do with how the camera handles the real world, but how us folk handle the resulting digital image. Its very similar as with video/tv, avoid striped shirts! Possibly, printing at some odd resolution may reduce the visibilty of the artifacts, say at 313 dots per inch may alter things slightly, possibly 433dpi may alter it more.

Since the individual sensor sites are arranged on a regular pattern, green probably being different to the others, then I could expect the colours to be seperated, as we saw.

Best wishes,

Ray

 

[StuartRae]

I don't know if anyone's looked at the embedded JPEG (320x240) from the DNG file, but having just had a look at Bart's very interesting article about down-sampling algorithms it does seem to confirm his theory.

It also reinforces my initial unease about the purple sky.

Here it is at 200% for better visibility.

Regards,

Stuart

 

[Asher Kelman]

Stuart,

This shows it is not just the distant lines that cause problems.

Also no machine makes marks like this by cutting straight lines into concrete! Not even in California with a tankful of speed!

So beware of sending your great fashion/newstory pictures as jpgs to any magazine editor!

What are the mathematical rules for minimizing this? Who has gotten rid of them with any method?

Asher

 

[ian sanderson]

Asher,

Just to let you know I downloaded the file. The moiré is quite evident and is cause for concern from my point of view as I am/was close to buying the M8. I wonder if the ISO 640 has got anything to do with it as noise mixed with the lines in the concrete might not help. I also noticed a lot of dust on the sensor, not wanting to change the subject but is the M8 easy to clean?

Kind regards

Ian