5D2 banding - business card

Started Jun 20, 2010 | Discussions
John Sheehy Forum Pro • Posts: 25,110
5D2 banding - business card

The subject has often come up, suggesting that the 5D2 banding is only present for "pixel peeping"; IOW, large image displays.

Here is a 5D2 ISO 100 blackframe downsampled to 300x200 pixels, about the size of a business card:

This is a histogram stretch, so the implied ISO push is rather high, but it serves to demonstrate how poorly banding disappears with downsampling, relative to 2-dimensional random noise.

-- hide signature --

John

gdanmitchell
gdanmitchell Veteran Member • Posts: 7,736
Re: 5D2 banding - business card

So, John, yet ANOTHER thread on this obsession?

zzzzzzzzz....

Dan

John Sheehy wrote:

The subject has often come up, suggesting that the 5D2 banding is only present for "pixel peeping"; IOW, large image displays.

Here is a 5D2 ISO 100 blackframe downsampled to 300x200 pixels, about the size of a business card:

This is a histogram stretch, so the implied ISO push is rather high, but it serves to demonstrate how poorly banding disappears with downsampling, relative to 2-dimensional random noise.

-- hide signature --

G Dan Mitchell - SF Bay Area, California, USA
Blog & Gallery: http://www.gdanmitchell.com/
Facebook: http://www.facebook.com/gdanmitchellphotography
Flickr: http://www.flickr.com/photos/gdanmitchell/
Twitter: http://twitter.com/gdanmitchell
IM: gdanmitchell

Gear List: Cup, spoon, chewing gum, old shoe laces, spare change, eyeballs, bag of nuts.

Peter Carmichael Contributing Member • Posts: 727
Re: 5D2 banding - business card

In the original raw, what DN are the brightest pattern noise features?

Legion5 Senior Member • Posts: 1,047
Re: 5D2 banding - business card

Have you tried turning off highlight tone recovery? Canon posted a bulletin saying that the majority of problems are because this feature is enabled by default.

Aperaturewise Contributing Member • Posts: 784
Re: 5D2 banding - business card

Geese.... I must have a bad monitor, all I see here is a beautiful misty green valley with a rainbow. Its a beautiful image

OP John Sheehy Forum Pro • Posts: 25,110
Re: 5D2 banding - business card

Peter Carmichael wrote:

In the original raw, what DN are the brightest pattern noise features?

Oh, a non-vandal poster! How nice!

The standard deviation of the original blackframe is 5.75. For the horizontal and vertical banding in isolation, it is 0.65 and 1.01. For a debanded version of the original, it is 5.66, not much less than the 5.75 statistically, but visually, the difference is tremendous. For the downsampled version, 0.40 ADU. The downsampled version of the debanded image is 0.27 ADU. The combined banding (hor and ver) of the downsample (with the debanded downsample subtracted) is 0.29 ADU. So, for the downsample, the banding in isolation is stronger than the 2-D random noise, statistically, but for the original, it is around 20% (it was more like 10%, BTW, with older Canons).

I didn't follow through with min and max values.

-- hide signature --

John

OP John Sheehy Forum Pro • Posts: 25,110
Re: 5D2 banding - business card

Legion5 wrote:

Have you tried turning off highlight tone recovery? Canon posted a bulletin saying that the majority of problems are because this feature is enabled by default.

I'll assume you're addressing me, even though you replied to Peter Carmichael. I am looking at the RAW data, completely outside the realm of HTP, which is just an implicit -1EC at half the gain.

-- hide signature --

John

Mescalamba Contributing Member • Posts: 886
Re: 5D2 banding - business card

Well, its interesting.. point is, can you prove it (show it) on some real-life photo? If its visible on regular photo, than it is problem..

Ive seen loads of photos from various cameras and 5DMK2 doesnt catch my eye with any problem (except fact I would expect bit better colors from that thing, but theres 1DsMK3 for that.. yea and resolution is, eh.. killed by AA filter).

OP John Sheehy Forum Pro • Posts: 25,110
Re: 5D2 banding - business card

Mescalamba wrote:

Well, its interesting.. point is, can you prove it (show it) on some real-life photo? If its visible on regular photo, than it is problem..

Prove what? Do you know what I asserted by posting this? I asserted that visible banding noise does not decrease as much as random read noise when you show an image at a small size.

Ive seen loads of photos from various cameras and 5DMK2 doesnt catch my eye with any problem (except fact I would expect bit better colors from that thing, but theres 1DsMK3 for that.. yea and resolution is, eh.. killed by AA filter).

The 5D2 has a weaker AA filter than is required for proper sampling. That's a shortcut for apparent sharpness, until cameras have many more MP.

-- hide signature --

John

fdxx Regular Member • Posts: 161
Appreciated

Although I don't push my images too hard to have any banding problems, I appreciate this and similar topics. If enough voices are raised, we will eventually get better cameras.
Thank you.
--
M.F.Demirhan

 fdxx's gear list:fdxx's gear list
Canon EOS 5D Mark II Canon EF 400mm f/5.6L USM Canon EF 17-40mm f/4.0L USM Canon TS-E 17mm f/4L Canon TS-E 90mm f/2.8 +6 more
UponQuai
UponQuai Contributing Member • Posts: 654
Re: Appreciated

fdxx wrote:

Although I don't push my images too hard to have any banding problems, I appreciate this and similar topics. If enough voices are raised, we will eventually get better cameras.
Thank you.
--
M.F.Demirhan

And I will switch to Canon if 5d2 or 5d3 is free from banding (at low iso ?)

A pure heart wish can not be treated as a troll, because I am serious

Earthlight Veteran Member • Posts: 3,209
Yep, it's easily visible in web sized images

And often seems to pop out more as the size is reduced.

Here's to hoping for better cameras in the future! And NO, this does NOT mean that the 5D isn't a very good camera for most shooting.

Earthlight

-- hide signature --
mattr Veteran Member • Posts: 3,532
Another example

ISO200 5D2 shot (shadows of course strongly pulled up) downsampled to 600x400, 1200x800, 2400x1600, and unchanged. Upper left is full frame, all others are 100% crops from lower right corner.

Peter Carmichael Contributing Member • Posts: 727
Re: Another example

Not sure where you're intending those examples to take us. Just like John's original post, it demonstrates that the patterns don't disappear from being sampled down to a small size, but you also demonstrate what happens when you drag elements of the capture from below the lower DR limit up to mid-grey.

As I say, not sure what you're trying to show us. Hope this isn't a blatant troll.

jsmiller Contributing Member • Posts: 826
Re: 5D2 banding - business card

John Sheehy wrote:

The subject has often come up, suggesting that the 5D2 banding is only present for "pixel peeping"; IOW, large image displays.

Here is a 5D2 ISO 100 blackframe downsampled to 300x200 pixels, about the size of a business card:

Let me sure I understand what you did in the downsample. Did you simply bin rows and columns to make the final image 300x200, or did you do something more sophisticated? Regardless, I would have expected any image the size of the 5D II binned down to this size to be extraordinary smooth, with just random noise, even with the log stretch. This is quite remarkable. Thanks for posting it. I assume the pattern is different on different exposures. This is a fixable fault, and Canon should fix it on future models, even if the vast majority of users will never know it is there. Some certainly will.

Joe

 jsmiller's gear list:jsmiller's gear list
Canon PowerShot ELPH 360 HS
mattr Veteran Member • Posts: 3,532
Re: Another example

Peter Carmichael wrote:

Not sure where you're intending those examples to take us. Just like John's original post, it demonstrates that the patterns don't disappear from being sampled down to a small size,

Yes, that's why I replied to John and called my post "Another example", not too difficult to figure out, right?

...but you also demonstrate what happens when you drag elements of the capture from below the lower DR limit up to mid-grey.

Yes, it demonstrates that banding limits the 5D2's useful dynamic range at low ISO, especially in downsampled images.

Come on Canon, other companies can do better in this regard. If you address this issue your very good cameras will become even better!

Peter Carmichael Contributing Member • Posts: 727
Re: Another example

mattr wrote:

Yes, that's why I replied to John and called my post "Another example", not too difficult to figure out, right?

My apologies. I tend to view the forum in "flat view", which means that it is not immediately apparent who is replying to who. You should bear that in mind when getting genuine and polite responses on an abuse-infested thread.

mattr Veteran Member • Posts: 3,532
Re: Another example

Peter Carmichael wrote:

My apologies. I tend to view the forum in "flat view", which means that it is not immediately apparent who is replying to who. You should bear that in mind when getting genuine and polite responses on an abuse-infested thread.

OK, good point.

furano Senior Member • Posts: 1,083
Re: 5D2 banding - business card

Mathematically the Mona Lisa is not perfect yet it is known as Art. People say their kids can do something like Picasso paintings, yet Picasso is an Artist. The Empire State Building has a lot of problems and wrong doings in construction, yet it is an iconic building.
In theory, EVDO can be faster than cable yet no cell phone is…

There are many incoherencies in theory vs reality and this is one… A billion pictures taken with a 5D2 but there is banding in black frames.

To be honest, I prefer to see the nice pictures than the black frames/candles/kitchen sinks and the such just to prove a point.

Personally I wish I could be paid to find problems in products or to have the amount of time used to find them, in that case, I would be happy taking more pictures.

Just my 2 cents…

-- hide signature --
Scott Eaton Senior Member • Posts: 2,224
Especially evident with tonal mapping

I don't have a 5D2, but I 've been borrowing a lot of cameras lately to see what I like the best.

I also do a lot of night shooting, and the past year or so have started enhancing my work with a bit of tonal mapping (not HDR). Tonal mapping often produces the same effect as artifically used here to produce the banding issue. Not sure if it's a coincedence but the banding pattern is curiously identical to Nikon 8000 scanning artifacts.

Over half the cameras I've used, including nikon, will exhibit banding issues like this if tonally stressed , although I have't used a D3 yet. The newer the camera, the more you tend to see it. You can induce it easily by simply taking a shot with the lens cap on and then tonal maping the image. I'm actually getting pretty good at running a de-noise filter on shadow areas to clean it up now because I'm so used to it given so many cameras I've used produce it.

I'm more curious at what causes this, and if specific exposure conditions poduce it (longer exposure -vs- shorter at the same E.V., etc). So, rather than bickering, the OP and the rest of the peanut gallery would be better off tryign to explain it and verifying conditions at which it occurs. Do two consequetive shots produce an identical pattern, do any camera parameters encourage the problem including seemingly idiotic things like using fresh batteries or a specific lens (circuit motor noise) for example?

Keyboard shortcuts:
FForum MMy threads