Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
06-02-2008, 11:39 PM   #1
New Member




Join Date: May 2008
Location: Utah
Posts: 7
K20D shadow noise pattern

I've been considering moving to the K20D from a different system, but there is something bothering me.

I have an acquaintance with a K20D that took it to Disneyworld not too long after getting it. On the "dark rides" (i.e. Pirates of the Carribean), he took full advantage of high ISO with his fast primes to get some really fun shots. By their nature, these types of shots have a lot of deep/clipped shadows (similar to say.. shooting a dimly lit concert) and they exhibit a strong noise pattern across the frame.

At a local level there is some large grain pattern in the shadows, but that's not the problem, it is the noise pattern across the entire frame that is the problem. At the vertical center of the frame (landscape composition) the noise levels are very good, but they get progressively worse toward the top and bottom of the frame. (and slightly worse at the bottom than the top)

I don't want him to get angry with me for posting his shots, so I'm just going to post vertical slices. The following are vertical slices of 2 different shots (NOT 100% crop, but rather resized) that he converted from RAW using Lightroom. I intentionally cropped to show areas of mostly just clipped shadows.

The left is ISO 4500 with a warmer white balance, and the right is ISO 2200 and white balanced to incandescent (blue channel torture test). I put them against a black background so the pattern stands out:



To be clear, the pattern comes from the noise only, these should mostly just be clipped shadows.

I had him send me a few of his RAW files, and the only way I've found to ensure consistent black levels throughout the frame is to clip the blacks to the highest luminance level of the shadow pattern (top and bottom of the frame), thus giving up a whole lot of dynamic range. Is this the only workaround?

Also, I'm wondering a few things: first, I've had a few members of this board send me RAW dark frames from their K20Ds, and I've also taken a few of my own at a local camera shop, and while I am seeing distinct patterns for each sensor tested, I can't produce a file with a pattern as pronounced his camera seems to have. This makes me wonder if he's got a bad sensor?

But then, they were also converted from RAW. Most of the high ISO K20D shots I've seen online so far have been camera JPGs, and they don't seem to have as much or any pattern visible. So does the JPG processing engine just clip the black level enough to make it invisible or is something else going on? Am I completely missing something else?

So.. how do your K20D shots that have a significant amount of deep shadows look? Particularly under incandescent light as the blue channel will always be the noisiest. In his shots, the pattern started to become obvious at ISO 1600 for incandescent, and nearer to 3200 for warmer white balances.

Thanks for your input...


Last edited by Code; 06-03-2008 at 09:16 AM.
06-02-2008, 11:54 PM   #2
Veteran Member




Join Date: Mar 2008
Location: Nowhere, Sweden
Photos: Gallery
Posts: 654
I took the liberty to make the noise more clear. To me thats either a faulty sensor or some major disturbance from the ride or something. Thats not ordinary noise, ordinary noise would never ever make a pattern like that.

If thats consistent on his camera, he should turn it in for repairs or a new one. I only have a K10D myself, but never have I seen anything like that.
Attached Images
 
06-03-2008, 12:18 AM   #3
Veteran Member




Join Date: Mar 2008
Location: Nowhere, Sweden
Photos: Gallery
Posts: 654
Here is a image I just took with my lense pushed down a pillow for darkness.

It was shot with ISO 1600 and brought into photoshop where I used levels to bring out all noise in it, even more then on your picture. I can not anywhere spot any of that uniform noise that your friend suffers from.
nn1 on Flickr - Photo Sharing!


And for completness here is the unedited version. Which is, surprise surprise, black.
nn2 on Flickr - Photo Sharing!

Btw, K10D is supposedly worse with noise, so my images are supposed to be worse
06-03-2008, 10:16 AM   #4
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
QuoteOriginally posted by Zewrak Quote
Thats not ordinary noise, ordinary noise would never ever make a pattern like that.
Exactly.

This is, I'm pretty sure, just remaining light from the scene. Invisible to the naked eye but still captured by the K20D

This is no lens cap shot and there are no reasons that the remaining light is evenly distributed. Note that the K20D captures 12Bit which exceeds your eyes capabilities. What looks black to your eye, may still contain lots of information.

If I look at your image at Gamma 0.7 (looks similiar to Zewrak's image while keeping a little bit more detail), I believe to see the opening of three adjecent tubes. Does this ring a bell?


Last edited by falconeye; 06-03-2008 at 10:24 AM.
06-03-2008, 10:36 AM   #5
New Member




Join Date: May 2008
Location: Utah
Posts: 7
Original Poster
The same pattern shows up on every shot, it is not light from the scene. The slice on the right (incandescent WB) shows some detail from the scene as there was no completely blacked out slice in the shot, but the color cast and luminance pattern are from noise.

I am very interested to see others' K20D high ISO shots with significant areas of deep/clipped shadow across the frame, particularly WB to incandescent light. (more than just dark frames) If you have camera JPG vs RAW conversion, even better.

Last edited by Code; 06-03-2008 at 10:56 AM.
06-03-2008, 12:24 PM   #6
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
QuoteOriginally posted by Code Quote
The same pattern shows up on every shot
If this were true, it would be a reason to get the camera replaced. However, you are not convincing me that this is not from background lumination (there obviously was light!). Even if in every shot, the (mostly invisible) ambient light may just have not changed as well...

Get a lens-cap shot and post it full-size here. See https://www.pentaxforums.com/forums/pentax-dslr-discussion/29046-how-post-unl...originals.html on how to do it.

Also, do the same test as carried out with dark-frame out-of-camera JPEG shots, cf. https://www.pentaxforums.com/forums/255501-post154.html. By comparison, you'll be able to tell if something strange is going on or not.

QuoteOriginally posted by Code Quote
I am very interested to see others' K20D high ISO shots with significant areas of deep/clipped shadow
@Code, I already provided this to you, in thread
https://www.pentaxforums.com/forums/pentax-dslr-discussion/28342-any-k20d-own...ling-help.html.

In the attachment, you find this shot again, with incandescent WB, and exposure-corrected by +5EV (read 25,600 ISO).
(#1) exposure-corrected by +4EV by ACR, then +1EV in PS,
(#2) +5EV in PS only,
(#3) with daylight AWB and +5EV in PS only.
(#4) +1.46EV in ACR, incandescent WB again, making it ISO2200 as in the original post.

Note that I do an EV correction of +1EV in PS by boosting the lower half of the histogram x2 to full. Because of Gamma, this isn't correct, actually. But a correct +5EV correction in PS is impossible because this would imply a boost 0f RGB=0.12 to 255...
So, the first sample would be all contained between RGB values 0 and 1 in the original! You may criticise the gradient. I find it amazing that this level of light amplification is possible at all! From what I know about sensor cell well capacities, we are talking about less than 100 electrons here!

Last edited by falconeye; 06-15-2011 at 05:28 AM.
06-03-2008, 12:52 PM   #7
New Member




Join Date: May 2008
Location: Utah
Posts: 7
Original Poster
QuoteOriginally posted by falconeye Quote
However, you are not convincing me that this is not from background lumination (there obviously was light!). Even if in every shot, the (mostly invisible) ambient light may just have not changed as well...
Every shot, taken on different days in different places/situations... same pattern.

QuoteOriginally posted by falconeye:
@Code, I already provided this to you, in thread
https://www.pentaxforums.com/forums/pentax-dslr-discussion/28342-any-k20d-own...ling-help.html.

Higher than ISO400 really doesn't do much more than shift the histogram and engage some extra noise suppression. With a non-broken sensor at least.
Yes you did, and thanks! I have it right here, WB set to the extreme left and boosted +3 EV so the pattern is obvious:



Your camera's pattern is slightly different (as are each of the 3 K20D's I have dark frames for) but they all generally have a lower level near the vertical center and higher near the top and bottom with a bit of horizontal banding.

It is obvious that the K20D sensor has a noise pattern, that question is moot to me.. I'm just trying to figure out at what point it ruins a typical incandescent lighting shot by answering a few questions:
1) Is the RAW converter to blame for making it so obviously visible? (I am thinking the answer to this one is yes)
2) Do in-camera JPGs exhibit the problem as badly as I am seeing it from RAWs? If so, what makes the difference? Is it the level at which blacks are clipped? The same shot with in-camera JPG next to a RAW conversion would really help.
3) Are some sensors just worse than others? I don't have a dark frame shot from the camera that took the original patterns I posted, but the other 3 I have are all about the same in terms of intensity... just can't compare them in a real situation since they are just dark frames.

I am moving away from a different system precisely because of noise pattern, so this is a very important consideration for me.

Last edited by Code; 06-03-2008 at 01:19 PM.
06-03-2008, 01:03 PM   #8
Veteran Member
*isteve's Avatar

Join Date: Nov 2006
Location: London, England
Posts: 1,187
Try a 40D, D60 and E510 and guess what - all of them have a feint pattern to high ISO shots when pushed. If you suppress it too much you also lose real information in the shot. Its something that happens with all sensors.

06-03-2008, 01:22 PM   #9
New Member




Join Date: May 2008
Location: Utah
Posts: 7
Original Poster
QuoteOriginally posted by *isteve Quote
Try a 40D, D60 and E510 and guess what - all of them have a feint pattern to high ISO shots when pushed. If you suppress it too much you also lose real information in the shot. Its something that happens with all sensors.
Yes, 40D has some banding, but it is of consistent intensity across the frame. Not sure about D60.

The E-510 has very bad, very structured horizontal banding, and the bands are in different, random places on each color channel. This just happens to be why I am switching systems (can ruin shots at ISO 800 + incandescent, even where shadows are not that deep)

Last edited by Code; 06-03-2008 at 01:54 PM.
06-03-2008, 01:33 PM   #10
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
QuoteOriginally posted by Code Quote
Yes you did, and thanks! I have it right here, WB set to the extreme left and boosted +3 EV [...] It is obvious that the K20D sensor has a noise pattern, that question is moot to me.. [...] I am moving away from a different system precisely because of noise pattern, so this is a very important consideration for me.
Ok, I see your point. BTW, I pimped my previous post a bit

One point of consideration is that the noise floor at ISO2200 at incandescent lighting shouldn't have been visible at all (cf. my last sample treatment).

The combination of noise pattern and noise amplitude is what must finally be judged. It isn't obvious that the dark-frame noise patterns will correspond to real noise patterns in images taken because dark-frame noise patterns measure the read-out noise which normally is dominated by the uniform shot-noise.

As for the difference between RAW and JPEG: JPEGs are clipped much heavier in the black. Cf. this infamous review page from Poland:
Test Pentax K20D - Szumy i jakosc obrazu w RAW - Optyczne.pl (scroll to the very bottom of the page).
06-03-2008, 02:07 PM   #11
Veteran Member




Join Date: Mar 2008
Location: Nowhere, Sweden
Photos: Gallery
Posts: 654
QuoteOriginally posted by falconeye Quote
If this were true, it would be a reason to get the camera replaced. However, you are not convincing me that this is not from background lumination (there obviously was light!). Even if in every shot, the (mostly invisible) ambient light may just have not changed as well...

-------------8<-------------<SNIP>
You get my APPROVED for a good post.
06-03-2008, 02:28 PM   #12
New Member




Join Date: May 2008
Location: Utah
Posts: 7
Original Poster
QuoteOriginally posted by falconeye Quote
As for the difference between RAW and JPEG: JPEGs are clipped much heavier in the black. Cf. this infamous review page from Poland:
Test Pentax K20D - Szumy i jakosc obrazu w RAW - Optyczne.pl (scroll to the very bottom of the page).
Thanks for that link... that table pretty much answers my question.. the pattern on that ISO 2200 shot in my original post is visible because the blacks aren't clipped high enough.. no bad sensor, etc...
06-03-2008, 06:07 PM   #13
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
QuoteOriginally posted by Code Quote
that table pretty much answers my question.. the pattern on that ISO 2200 shot in my original post is visible [...]
Great! I never understood that table. What does it say in non-Polish words?

The sample which I provided shows that at ISO 2200 and 1/30s, a K20D black frame developed from RAW (no clipping of Blacks at all (=0)!) turns out (almost) black, with hardly anything visible to the naked eye -- even with incandescent WB. Your sample showed a clearly visible dark blue background.

So either, you boosted the shadows in PP, or the exposure was a lot longer than 1/30s (I doubt it because then the long exposure NR kicks in...), or there was ambient light (what I assume there was!), or something strange is going on.

I provide yet another real-world example for your convenience.

The photo of a candlelight diner table flower at 2.8LV incandescent luminosity.
Photographed with K20D/RAW at ISO1600 and color-corrected for 2100K (strong incandescent WB!). Black clipped at 0 (for both photos)!

The second got a true +4EV correction in LR!! (With restored highlights) You see that the black upper spot in the upper left corner comes out fine while the black hairs show banding. But this is banding at effective ISO 25600! It is amazing that there was structure in the black hairs to recover at ISO1600 at all...

What I wanted to say: your non-uniform noise distribution is really of no concern here. Even banding is quite well under control (but it is not zero). One should probably average the very blackest areas for optimum results. Maybe, some noise reduction tools will do just that. This is probably as good as it gets with the K20D. Your turn to decide if this is what your are looking for.

Last edited by falconeye; 06-15-2011 at 05:28 AM.
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
bottom, camera, dslr, frame, iso, k20d, level, noise, pattern, photography, shadows, shots
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Example of nasty banding (pattern noise) with Pentax K20D mr.twister Pentax DSLR Discussion 3 09-03-2009 07:01 PM
Is this noise pattern normal? pre86 Pentax DSLR Discussion 8 08-18-2009 11:17 AM
Strange K10D noise pattern miriya Pentax DSLR Discussion 6 03-15-2009 10:09 AM
K20D - pattern noise: show me your worst please chse Pentax DSLR Discussion 11 10-02-2008 03:20 PM
K10D and weird noise pattern quickpix Pentax DSLR Discussion 14 01-28-2008 03:56 PM



All times are GMT -7. The time now is 10:17 PM. | See also: NikonForums.com, part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top