Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
10-16-2010, 06:53 PM   #46
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
QuoteOriginally posted by jpzk Quote
I just want to find out at which ISO limit do you guys think that NO NR would be necessary (with properly exposed shots)?
No NR may not be the most meaningful benchmark. Automated NR as in LR3 (or any automated NR applied to unprocessed raw data) is state of the art. You don't dismiss hot pixel removal or Bayer pattern extrapolation in demosaicing either.

But if you really really want to know at which ISO noise starts to degrade IQ (when unprocessed) then DxO seems to think around 30dB SNR for an 8MP image. Currently, this limit is already reached below ISO 900 for APSC and below ISO 2000 for FF. If the K-5 is excellent in this department (which remains to be seen) than it would reach this limit at ISO 1000. Of course, most photographers accept lower than 30dB SNR. With good NR software, this limit drops to about 24dB which corresponds to about ISO 3200 for APSC (ISO 7000+ for FF). Pentax applies NR to its raw file and therefore achieves ISO 5000+. Without tricks (ie., pushing ISO 1600 only), then ISO 4000 would be a very good result for the K-5 in this department. Beyond, some image degradation even with NR software will be visible. E.g., I still can see a softening effect in the upper glass texture in JohnBee's 6400 sample.

Regarding NR: it doesn't kill image detail if done with a sophisticated algorithm (I've written one such algorithm and image detail like text can be preserved w/o softening it). State-of-the-art NR shows itself in noise-like textures which actually look like noise and therefore, are destroyed. E.g., my algorithm assesses the amount of statistical likely noise and tries not to remove more than this expected amount, leading to natural results with proper calibration. This is why I looked at the glasses' top border which has texture from dirt.

10-16-2010, 07:43 PM   #47
Veteran Member
JohnBee's Avatar

Join Date: Sep 2008
Location: Newrfoundland
Photos: Albums
Posts: 4,667
Original Poster
QuoteOriginally posted by falconeye Quote
E.g., I still can see a softening effect in the upper glass texture in JohnBee's 6400 sample.
What is your take 'if'... this was based from a RAW rather than JPG?
Keeping in mind, that JPG is less profitable than RAW.
I would expect both samples to be too close to call at 100%.

The only advantage I can see with the D700 comes with respect to low level NR.
However... it appears that the overhead in the K-5 files may compensate for this also.
Though from what I can see, low level NR doesn't seem to impact IQ to the point of degradation any further than the D700.

Also... I'd be curious to see how things would hold up from an ISO1600 file pushed into the same levels(6400). Purely for the sake of avoiding low level NR.

Having said that, I can't get any better using a D700 ISO6400 RAW either.
And I'd like to think that I've spent enough time with the D700 files, to make a worthy assessment of that.

What we really need here, are some K-5 DNG/RAW files...
Can't you just snap your fingers or something?
10-16-2010, 07:55 PM   #48
Veteran Member




Join Date: Dec 2007
Posts: 923
NO NR is useful for comparing camera output, but I never dismiss low-level NR in actual usage.
For low-to-moderate level High-ISO noise, just a light touch of NR can make a very pleasant improvement to the visual quality of a photograph.
Good NR software will do it with minimal , almost unoticeable loss of detail. I use Noiseware Pro - good stuff.
10-16-2010, 10:59 PM   #49
Veteran Member




Join Date: Dec 2007
Photos: Gallery
Posts: 8,237
NR software seems to be what's on display here, though...

I'm really loving these k-5 examples, but I guess I would like to see some untouched-by-NR-in-PP samples as a benchmark. I've see some Nikon D5000 samples run through Topaz Denoise that looked stunning also, and we know that sensor is good, not great (by current standards.)

Another thing that nags at me is time - on my laptop (where I do 90% of my PP, because it's always with me,) running a typical shooting session through Topaz Denoise takes about an hour. My time is not worth $0/hr. If I add up that time, after a couple months, it comes out to the difference between say a D700 and a K-5.

(Of course, I guess I could multitask better and get that time back )

10-17-2010, 05:01 AM   #50
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
QuoteOriginally posted by JohnBee Quote
I would expect both samples to be too close to call at 100%.
Actually, your "case study" already provides two samples being too close to call. I had to try hard to spot the minimal difference. Good job.

My point was that a minimal loss of information is unavoidable at ISO 6400 even with a camera as stunning as the K-5. After all, it's only 1/64 of photons which are <1000 per pixel reaching the sensor (actually, it's about 6000 which ~600 of are detected after the Bayer filter) ... NR may hide most of this. But not all.

I provided an ISO 6400 raw sample early on and it has textured areas (the printing pattern) which start to smear out and the NR can't do anything about it.

I compared a K-x and a D700 (both 12MP) in the lab. The D700 has 1 stop less noise at ISO 1600 (Falk Lumo: Lumolabs: Sensors of Nikon D700, D5000 and Pentax K-x) and appearantly about equal noise at ISO 12800. However, the noise then looks more smeared for the K-x (D700 has more fine grain noise looking more pleasing) and DxO confirms (by a cross correlation test) that the K-x plays tricks with the raw data beyond ISO 1600. My article contains test patches you can look at. The Pentax tricks may confuse the average tester but not DxO, .. or me

The final call is still out. But for the moment, I have no reason to believe that the quantum efficiency of the 16MP Exmor HD is higher than that of the 12MP one in the K-x. Therefore, the resolution-normalized 18% gray noise measurements will only improve that much. You can't beat the laws of physics.

Where the K-x, and probably K-5 in particular, does excell is read-out noise rather than quantum efficiency. It is probably low enough to make a difference at ISO 100,000+ to challenge the D700. But not below. Where the read-out noise shows off though is dynamic range below ISO 200 where it will probably beat the D700.

Don't misunderstand me: the K-5 has a stunning sensor, esp. wrt DR and as it seems, color rendition.

But it does nothing about the 1 stop difference between APSC and FF in high ISO noise. And the current king in high ISO noise (the D3s) even stays almost 2 stops away ... If the K-5 can narrow this gap then that would be great



Just don't say K-5 gives FF a run for its money. If you do, you pay with my voice of ratio
10-17-2010, 05:23 AM   #51
Site Supporter
Site Supporter
Digitalis's Avatar

Join Date: Mar 2009
Location: Melbourne, Victoria
Photos: Gallery
Posts: 11,694
when looking at images with Noise reduction applied don't look at the detail look at the Out of focus portions, the smooth gradations are the most difficult for NR programs to deal with because Low frequency noise plays merry hell with such subtle gradations. NR usually gives it the appearance of lumpy porridge, especially if you are using a lens with bad bokeh(which isn't the case here). That's how I can spot the difference between the two images the OP posted.
10-17-2010, 01:37 PM   #52
Veteran Member
JohnBee's Avatar

Join Date: Sep 2008
Location: Newrfoundland
Photos: Albums
Posts: 4,667
Original Poster
QuoteOriginally posted by falconeye Quote
I provided an ISO 6400 raw sample early on and it has textured areas (the printing pattern) which start to smear out and the NR can't do anything about it.
Do you mean this sample?



Though slightly overprocessed, I find the printing pattern to of come through very well from my end-of things. However, without a nominal sample from which to draw a contrast with, it is very hard to make an accurate assessment.

A notable thing with this image is where the image is pushed(nearly 1 stop) in ACR prior to processing. Which imo. serves as a positive toward the integrity of the file since noise and detail should improve with a properly exposed image also.

Let me know if you think this makes any sense.


Last edited by JohnBee; 05-01-2011 at 10:46 AM.
10-17-2010, 02:48 PM   #53
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
QuoteOriginally posted by JohnBee Quote
Do you mean this sample?
Yes, I know you looked at it before

At the lower right corner, the forth-most dark printing pattern (the last one with a visible texture), smears out a bit at ISO 6400. Nothing to be worried about. Its only to illustrate that ISO 6400 isn't ISO 80
10-17-2010, 03:15 PM   #54
Veteran Member




Join Date: Apr 2009
Location: Dallas, TX
Photos: Gallery | Albums
Posts: 499
QuoteOriginally posted by kittykat46 Quote
Similar process with Noiseware.
Just default, plain vanilla Noise reduction. You can process a whole bunch of shots in batch.

With my K-7 , Noiseware default NR just didn't cut it on ISO 3200 onwards - I had to learn to work all the sliders left and right.

I think that level of "skilled NR" would only be needed on ISO 25600 - 51200 for the K-5, hard to imagine when I would use THAT.
Ummm...Midnight, no moon, f8 500mm lens trying to take action shots of owls or bats with no flash or other lighting? Man...I cringe at anything over ISO 400 on my K10D. I'm still waiting for the 'new toy' period to get over, some actual reviews, and the price to drop...but from what I've seen thus far it's looking like there's a K-5 in my future. Let's hope everyone who has one loves it as much as the few I've read of so far.
10-17-2010, 03:34 PM   #55
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
QuoteOriginally posted by opiet70 Quote
action shots of [...] bats with no flash or other lighting?
I heard that the K-5 can activate its ultrasonic dust shaker to create an ultrasonic sound field and then takes great bat action shots from the sound reflections. I the total dark! In bat scene mode, the camera even tries to emulate a genuine bat sound field to appear friendly to other bats

I wanted to try it. But I couldn't find the menu setting to activate it in the total dark
10-17-2010, 05:22 PM   #56
Veteran Member




Join Date: Jun 2009
Photos: Albums
Posts: 1,674
QuoteOriginally posted by falconeye Quote
I heard that the K-5 can activate its ultrasonic dust shaker to create an ultrasonic sound field and then takes great bat action shots from the sound reflections. I the total dark! In bat scene mode, the camera even tries to emulate a genuine bat sound field to appear friendly to other bats

I wanted to try it. But I couldn't find the menu setting to activate it in the total dark
first you have to get bitten by a radioactive spider...after that it gets complicated.
10-18-2010, 07:10 AM   #57
Veteran Member
JohnBee's Avatar

Join Date: Sep 2008
Location: Newrfoundland
Photos: Albums
Posts: 4,667
Original Poster
QuoteOriginally posted by falconeye Quote
Yes, I know you looked at it before

At the lower right corner, the forth-most dark printing pattern (the last one with a visible texture), smears out a bit at ISO 6400. Nothing to be worried about. Its only to illustrate that ISO 6400 isn't ISO 80
Do you mean this one?



I find it easier to assess final IQ at 200%. though I thought it looked really good.
However without a nominal sample to contrast with, there was simply no way to tell how well the image was holding up.

Would you happen to have a lower ISO DNG or JPG of this scene?
10-18-2010, 07:21 AM   #58
Senior Member




Join Date: Oct 2008
Posts: 213
Oleg_V from dpreview said: " K-5 at ISO100 the NR strength is about 2 times larger than at ISO6400 (!). In numerical values it is 26% (green channel) or 2/3 of the ISO stop. In terms of the 'effective resolution' its only 10Mpix.

don't understand why to use NR at ISO100 ?
10-18-2010, 07:28 AM   #59
Veteran Member




Join Date: Dec 2007
Posts: 923
QuoteOriginally posted by Vaikis_ Quote
Oleg_V from dpreview said: " K-5 at ISO100 the NR strength is about 2 times larger than at ISO6400 (!). In numerical values it is 26% (green channel) or 2/3 of the ISO stop. In terms of the 'effective resolution' its only 10Mpix.

don't understand why to use NR at ISO100 ?
He doesn't have a K-5 in his hands.
I've learnt to take everything he writes with a heavy pinch of NaCl
10-18-2010, 07:37 AM   #60
Senior Member




Join Date: Oct 2008
Posts: 213
QuoteOriginally posted by kittykat46 Quote
He doesn't have a K-5 in his hands.
I've learnt to take everything he writes with a heavy pinch of NaCl
I think that he made a mistake
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!
camera, dslr, iso, iso6400, iso80, k-5, k-5 ii, k-5 iis, k5, pentax k-5, people, photo
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
K-7 ISO3200 Advanced NR Case Study JohnBee Digital Processing, Software, and Printing 3 06-18-2010 01:02 AM
K20D ISO6400 Image set JohnBee Pentax DSLR Discussion 21 05-16-2010 12:24 PM
Lets try some ISO6400 with LR3 Beta... JohnBee Pentax DSLR Discussion 26 04-20-2010 05:17 PM
K-x iso6400 black cat planedriver Pentax DSLR Discussion 4 12-15-2009 08:28 AM
For Sale - Sold: Early Takumar 300mm f/4 Excellent Cond. Case, Caps heliphoto Sold Items 10 03-06-2009 09:31 AM



All times are GMT -7. The time now is 07:12 AM. | See also: NikonForums.com, CanonForums.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