aliasing errors?

[robertwright]

Advertisement (gone after registration)

I have been getting this on backlit subjects, wondered what everyone thought. Interesting that it is not everywhere on overexposed backlit subjects. I have other examples, often if I photograph someone backlit I will get it on a shoulder. I have assumed that it was due to the lack of an AA filter, the camera just is making wrong decisions on the demosaicing. I usually have to go in and fix it with a blur or desaturate.

 

Before anyone says too much over exposure, I agree, it is harsh, but not unusual to overexpose in many situations, and as this example shows, it only happens on small areas, not all over the overexposed area. I also get it on lamps in frame sometimes, just along the edges.

 

100% crop, lens is elmarit 21 third version (cdn), prolly f8. but it happens on my 35cron also, I would assume on any lens. also it is independent of raw software.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[robertwright]

here is another example, 35 cron at f2, 320iso.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[bill vann]

have looked at my images and don't see a similar phenomenon?

 

don't have anything quite so contrasty that i've found yet but what i have reviewed doesn't show this.

 

kindly

 

bill

[mike prevette]

This effect is a trate of the sensor not the AA filter. Although the AA filter would probally reduce the noticible effect just based on the fact it reduces edge contrast. Remember a sensor is actually an analog device that gets read out in sections. A "high" signal next to a "low" signal can effect eachother and screw up how the bayer demuxing is computed.

 

_mike

[robertwright]

well it is interesting that it is not all over the transition edge in these examples, it is in some places and not others. I wonder if firmware could have an effect on this, in other words, whatever algorithm is being used to "decide" needs some education to make better decisions.

[carstenw]

This has been documented in some previous threads, and happened under similar circumstances. I hope Leica fixes this one. It doesn't happen very often, but when it does, it is tricky to eliminate.

[robertwright]

Advertisement (gone after registration)

This has been documented in some previous threads, and happened under similar circumstances. I hope Leica fixes this one. It doesn't happen very often, but when it does, it is tricky to eliminate.

 

I see it very often, if there is a light in the frame usually one one edge I will get a little of this. I guess I have to pray Guy takes it up the mountain to the gods to decide...

[sandymc]

This effect is a trate of the sensor not the AA filter. Although the AA filter would probally reduce the noticible effect just based on the fact it reduces edge contrast. Remember a sensor is actually an analog device that gets read out in sections. A "high" signal next to a "low" signal can effect eachother and screw up how the bayer demuxing is computed.

 

_mike

 

Sorry, but no, this has nothing to do with the sensor - the Kodak sensor has protection against blooming, etc. This is straight aliasing - the reason why you tend to see it on transitions from very bright areas is just because of the amount of light present. If the adjacent pixel is several stops above white, even a small amount of aliasing will be very visible. If the adjacent pixel is a mid tone, the aliasing is still there, but 5-6 stops less bright, so nearly invisible.

 

BTW, this issue is (was) well known on the old Kodak digital cameras, which also had no anti-aliasing filter - search on "purple fringe"

 

Sandy

[Guest sirvine]

Huh...I never noticed it, but I've been resisting the urge to pixel peep on the specular highlights that are bothering me in M8 shots. I used to do a lot of video, and sometimes the highlight clipping in M8 shots reminds me of digital video. The kind of artifact you've got here would drive me batty.

[carstenw]

Sandy, this is not normal aliasing. You can see maze artifacts in the border region sometimes.

[tashley]

My first M8 has never done this. When I bought my second, this effect was visible around many shots of transitional brightness, eg sunsets, around the sun, or shots with bright light sources like halogen bulbs. I returned the second body immediately and got a replacement, no questions asked, and that body does not do this.

 

In my book, that makes it at best an 'occasional trait of the sensor' - or at least the way in which it is processing data - but in reality it is an error and you should ask for a repair or replacement.

 

Tim

[jaapv]

I wonder what film would do in these conditions.

[robertwright]

Does anyone else think this is due to faulty hardware?

I have two of the same pictures with different exposures in the bracket for the first example, the posted pic is 1/2 second, but at 1/4 second it is not evident, however, if I boost the exposure in post I can see it.

Also it happens regardless of the amount of sharpening applied, assuming that 0 sharpening in Lr is indeed 0.

Again my suspicion is that this is a software problem-as for film, film would simply render the flare around the object in a continuous manner, being an analogue medium. Scanning and printing it could result in some sort of posterization for example, but usually this would happen in shadows not highlights.

 

It would be willing to send examples to the Leica engineers if this would help identify the issue. But really I would need to do a controlled test to determine what level of overexposure produces this, and whether or not other factors are involved.

 

I have also seen interesting artifacts on window blinds, a kind of pixel bleeding around the edges of the window where light should not be, but shaped like the slats of the blind. I also have one example of a man's head with blind shaped cutouts caused by an over exposed window blind behind him.

 

Not saying that the camera should necessarily be able to traverse this, I do understand that without the AA filter we are "seeing" the chip in all its hoary glory, which usually I like. But I can't think of a similar problem on a canon camera for example, which I used to use all the time before the Leica.

 

Overall it seems the Leica requires something of the opposite approach from other digitals, on other digitals a little over exposure is required to keep the midtones from sliding down, on the Leica, any over exposure is deadly either in artifacts or highlight compression artifacts, but the compensation is that the midtones are pretty elastic.

 

I find myself checking the histogram and backing off any hint of overexposure.

 

opinions?

[sps]

I have not had this problem with mine. You might need a trip to the service dep't.

[tashley]

All I can add is that when I posted my sample shots, the general opinion was either that I was asking for trouble by photographing bright things, or that it was a software issue.

 

I kind of knew this was wrong from the obvious fact that my existing camera did not do this, even under identical circumstances. That's why I felt justified in asking for an immediate exchange. Take a close look at this thread and see if it describes your problem:

 

http://www.leica-camera-user.com/digital-forum/18949-end-world-we-know.html

 

If it does, then the camera should not be doing this. It is NOT common to all M8s, it is an error, and you should get it dealt with.

 

Tim

[robertwright]

well if your melty sun had the pixel maze then I would say it is the same problem, but I cannot tell from the posted crop. There was another images of a lamp there with the pixel maze, same problem.

 

I'm not wholy convinced it is a hardware problem, if only that it is so random in similar circumstances. My vote is that it has to do with the bayer interpolation algorithm. Is there a way to approach Leica support via email to send samples?

[sandymc]

 

Umm, well, if you're shooting raw, which your first post suggests you are, then the Bayer interpolation code is in whatever raw processor you're using e.g., C1 or Lightroom, none of which have anything to do with Leica.

 

Sandy

[gdeliz]

It looks like simple aliasing to me. Most shots taken with the M8 or any DSLR without an AA filter will have aliasing whether it is noticeable as such or not. The degree of aliasing will be dependent on lens and aperture as well as subject matter. The sharper the lens, the more aliasing and a lens at its optimum aperture will show more aliasing than a lens wide open. On the M8 the sensor will start to become diffraction limited at about f/11 so stopping down to f/16 may eliminate the worst of the aliasing in problematic images. Someone might want to run some tests.

George Deliz

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[Riley]

George maybe right

or is it, by George youve got it

 

The Sigma SD14 likewise has no AA filter being foveon equiped

and seems to have problems with some RAW converters

ACR being just one of them

[robertwright]

thanks for the responses,

 

it is not dependent on raw software, it happens with all converters. This was my first guess when I saw it, was to run it through C1 but it was the same.

 

interesting response about the sharper the lens, on the sample with the man, that was the 35 asph cron wide open, which is pretty sharp already at f2. The other sample is the elmarit 21 (cdn) at 5.6, which I think is the optimum aperture. I did a looksee around other images for similar effects and noticed that the elmarit wide open is resistant to this, meaning, it is not so sharp wide open, which I know to be true.

 

I think this means I could do a test and vary aperture to see if it helps, I think diffraction will soften it for sure.

 

Next time I am out on a job with the camera on a tripod shooting against a window, I will shoot a full bracket to see if it goes away.

 

conclusion might be sharp lenses at optimum apertures and no AA filter can produce this in situations with hard transitions. If this is true it might be iso independent.