M8 vs DMR, very different colors in Capture One

[pascal_meheut]

Advertisement (gone after registration)

One another thing: I did a quick test and shot a very bright and small light source in front of a white wall at different exposures.

Banding occurs at 320 ISO, 640 and 1250 but not at 160. Did not tried 2500. It is more or less visible depending on the sensitivity, the exposure...

But I also did the same shot with the DMR and no banding at all at 800 ISO and even at 1600 ISO. Well, at 1600 the image show banding oll over it but not because of the light source.

 

Does someone knows why such close parents CCDs can give such different results ?

[rvaubel]

One another thing: I did a quick test and shot a very bright and small light source in front of a white wall at different exposures.

Banding occurs at 320 ISO, 640 and 1250 but not at 160.

 

Does someone knows why such close parents CCDs can give such different results ?

 

Well, one major difference is the bit depth of the software. The M8 is reported by most to have only 8bit non lineal DNG file. Could this result in a banding artifact? One thing that leads one to believe it could be true is the "white on lite" effect. In a non-linear scheme the highlight readings may not have enough values to cope with a blown source.

 

Rex

[scott kirkpatrick]

Let's use "bleeding" to describe the horizontal streak that comes from an extra bright source and spreads horizontally across the image from right to left. That's a better term than "banding," which (I think) is usually applied to posterization of shadow or highlight tones, and doesn't have to be horizontal. The bleeding effect doesn't seem to be seen in the DMR, even among those who like to show off their flare-free Leica lenses by including lights in the image.

 

The chips in the DMR and the M8 are nominally the same product (KAF10500), but since the M8 has offset microlenses and its batch of chips was manufactured almost a year later, there could certainly be differences in processing parameters that were thought not to matter.

 

To quantify this, you need to report the brightness of the spot and the ISO setting. A bare light bulb is too hard to measure, but a lamp globe gives a reasonably useful spot meter reading. Presumably the problem has something to do with generating so many electrons per pixel (the limit is 60,000) in the image of the bright spot that the mechanisms designed into the chip for getting rid of excess charge are overwhelmed, and the excess spills down the row into adjoining cells as the image is read out.

 

Here's a thought -- find a scene in which bleeding occurs. Then shoot a few more times, each time throwing it further out of focus to spread the bright spot, and see when the bleeding stops.

 

scott

[scott kirkpatrick]

Well, one major difference is the bit depth of the software. The M8 is reported by most to have only 8bit non lineal DNG file. Could this result in a banding artifact? One thing that leads one to believe it could be true is the "white on lite" effect. In a non-linear scheme the highlight readings may not have enough values to cope with a blown source.

 

Rex

 

But in the M8 dng encoding, each pixel is encoded (by taking its square root to leave only 8 bits) independently. This shouldn't spread the effect of a hot group of pixels over the pixels downstream, at least it would take a pretty strange sort of firmware bug to make an overflow error spread out in that way.

 

scott

[rvaubel]

But in the M8 dng encoding, each pixel is encoded (by taking its square root to leave only 8 bits) independently. This shouldn't spread the effect of a hot group of pixels over the pixels downstream, at least it would take a pretty strange sort of firmware bug to make an overflow error spread out in that way.

 

scott

 

I agree that the blown highlight initiates the effect,"bleeding" if you will. The weird part is that a simple overflow should effect only the adjacent pixels and be mitigated rather quickly. But the artifact manifists itself as a more or less continous tone "band" across only the horizontal plane. And only on a light colored background. Blooming/overflow effects should show up on dark backgrounds too.

 

There is something that looks sort of software generated about it but I can't put my finger on it.

 

Rex

[gepetto]

Pascal what color space are you setting the M8 to capture? Could the C1 M8 profile be calibrated for using a particular color space? This could cause a color shift if they are not matched correctly.

 

Interestingly for Mac users the ColorSync Utility has a generic profile for the Leica M8 and also the R9/8.

[pascal_meheut]

Advertisement (gone after registration)

No, this is not a profile issue. Because the setting on the camera changes nothing when shooting raw and because I tried several combination to be sure.

 

But according to a friend, a Nikon expert, the D200 had the same "banding" or "peeking" a problem but not all the cameras, only some. So I'm interested to know if other M8 users see it on their pictures or not.

[robsteve]

Pascal:

 

Edmund Ronald just posted a query about it on the Fred Miranda Forum. He just got his M8 and is also in Paris.

 

•Hands-On• Leica M8 - FM Forums

[robsteve]

Pascal:

 

I just tried runnning your picture through Band Aide by Camera Bits. It was a photoshop plugin designed when the early Nikon D1 had very bad banding problems at 800ios and higher. The Canon 1D also had this banding, but a firmware updare made it much better.

 

It seemed to work very well on the banding in your sample image. I had used the weak setting and 50% You might want to download the demo and try it on the full size file.

 

I adjusted the color a bit too.

 

Welcome to Camera Bits, Inc.

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!

[JeffWright]

Pascal:

 

I just tried runnning your picture through Band Aide by Camera Bits. It was a photoshop plugin designed when the early Nikon D1 had very bad banding problems at 800ios and higher. The Canon 1D also had this banding, but a firmware updare made it much better.

 

It seemed to work very well on the banding in your sample image. I had used the weak setting and 50% You might want to download the demo and try it on the full size file.

 

I adjusted the color a bit too.

 

Welcome to Camera Bits, Inc.

 

That did make a big difference, but you can still see the bands. Did you do some noise reduction to, Rob?

 

Glad to see all of this discussion here...the leica-camera forum has taken off since I last visited. Probably a good thing since I may have just gotten myself banned from FM....

[robsteve]

All I did was use curves to adjust the color a bit, run the Band Aide Plugin, and then do a smart sharpen.

 

I need a sharper sample and a full DNG to play with to see how much the Band Aide plugin effects fine details

 

Robert

[pascal_meheut]

Thanks for all the infos, Robert. You can download the DNG here:

 

http://www.photofaq.eu/temp/L1000401.DNG

 

I'll try Band Aide myself too.

[sean_reid]

I think Pascal may be sleeping at the moment given that he's in Europe.

 

I thought of that when I posted but I figured he'd see it in the morning.

 

Sean

[pascal_meheut]

Sean, I saw it and answered your questions.

[scott kirkpatrick]

The man in the cafe (by RPO) was shot at ISO 1250 and 1/250 sec. The amount of bleed is light, but it is over a light background (nearly filled pixels don't soak up excess charge very fast) and extends for maybe 400 pixels, only visible downstream (to the left).

 

The Printemps shot is rotated, but the auto headlights bleed a little bit upstream and to the edge of the frame downstream -- perhaps 800 pixels, showing very lightly over a dark background. Downstream is down in this picture. This was shot at 1/90 and ISO 1250.

 

The park bench shot shows the greatest stretch. It was shot at 1250 and 1/8 sec. The streetlights either bleed all the way across the frame, over a dark sky, or more likely, bleed both upstream and downstream for a shorter range.

 

So a slow shutter speed allows this to occur more strongly, making it seem that the excess electrons spill over during exposure.

 

scott

[roblumba]

Are these shutter speeds slow enough that the black frame subtraction technique is being used by the camera? I wonder if the subtraction of a black frame is contributing to this?

[pascal_meheut]

No, I can produce some at any speed.

[scott kirkpatrick]

Are these shutter speeds slow enough that the black frame subtraction technique is being used by the camera? I wonder if the subtraction of a black frame is contributing to this?

 

I thought I read somewhere that the M8 does dark frame subtraction at and below 1/30, so only the last shot would involve it. Besides, it is intended to subtract a different kind of noise due to leakage from the silicon substrate. Shouldn't have any effect here.

 

scott

[Guest xxl-user]

for me m8 banding is no problem.

 

most important thing is the red dot on this camera.

 

arnold

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[pascal_meheut]

for me m8 banding is no problem.

most important thing is the red dot on this camera.

 

Quite the contrary, I put some black tape on the red dot to hide it