| Moderated by: chrisbet, |
|
|
| High iso noise "blind test" | Rate Topic |
| Author | Post |
|---|
| Posted by Gert: Mon Jun 18th, 2012 14:22 | 1st Post |
| I just got a new camera and decided to make a noise comparison with the old one, they are very close, so would like a second opinion. To get an unbiased opinion, I am not going to tell you what models it is (don't worry it is 2 Nikon cameras), so that way it is a "blind test". First, some test background and then (next post) some pictures. Since noise at a certain ISO is not always the same, but incises as the shutter time goes up, as well as light quality goes down (WB corrections will increase noise in the other colour channels), this test is done with just one fluorescent lamp. Both cameras had a custom WB, and was manually focused using live-view. All pictures was taken with the camera on a tripod and mirror-up. The raw files was processed into jpgs with Aftershot Pro (formerly known as Bibble). Then crops was made. The shutter speed ranges from 2.5 sec @ iso 200 to 1/13 sec @ iso 6400, so this is a fairly cruel test. First is a down-sampled picture of the scenery with daylight WB, so give you an idea of the setup. Attachment: resize.jpg (Downloaded 49 times)
|
| Posted by Gert: Mon Jun 18th, 2012 14:23 | 2nd Post |
| To show the best possible, we start with 2 pictures at iso 200 Camera A Attachment: A-200.jpg (Downloaded 48 times)
|
| Posted by Gert: Mon Jun 18th, 2012 14:25 | 3rd Post |
| Camera B @ iso 200 Attachment: B-200.jpg (Downloaded 46 times)
|
| Posted by Gert: Mon Jun 18th, 2012 14:27 | 4th Post |
| Camera A at iso 1600 Attachment: A-1600.jpg (Downloaded 46 times)
|
| Posted by Gert: Mon Jun 18th, 2012 14:28 | 5th Post |
| Camera B at iso 1600 Attachment: B-1600.jpg (Downloaded 46 times)
|
| Posted by Gert: Mon Jun 18th, 2012 14:29 | 6th Post |
| Camera A at iso 3200 Attachment: A-3200.jpg (Downloaded 45 times)
|
| Posted by Gert: Mon Jun 18th, 2012 14:29 | 7th Post |
| Camera B at iso 3200 Attachment: B-3200.jpg (Downloaded 45 times)
|
| Posted by Gert: Mon Jun 18th, 2012 14:30 | 8th Post |
| Camera A at iso 6400 Attachment: A-6400.jpg (Downloaded 45 times)
|
| Posted by Gert: Mon Jun 18th, 2012 14:30 | 9th Post |
| Camera B at iso 6400 Attachment: B-6400.jpg (Downloaded 45 times)
|
| Posted by Gert: Mon Jun 18th, 2012 14:32 | 10th Post |
| Please let me know which one you thing is best - if any - and why. I will let you know what models it is later on. Thanks, Gert
|
| Posted by Ed Matusik: Mon Jun 18th, 2012 15:44 | 11th Post |
| To my unpractised eye, Camera B would be my choice. "B" doesn't show really unacceptable noise until ISO 6400. - EdM
|
| Posted by TomOC: Mon Jun 18th, 2012 17:19 | 12th Post |
| I agree with Ed. Camera B is a clear winner, even at lower ISO Tom
____________________ Tom O'Connell -Lots of people talk to animals.... Not very many listen, though.... That's the problem. Benjamin Hoff, The Tao of Pooh |
| Posted by jk: Mon Jun 18th, 2012 17:38 | 13th Post |
| Yes camera B as well for me! Now dont say it is a Leica M9.
____________________ Still learning after all these years! https://nikondslr.uk/gallery_view.php?user=2&folderid=none |
| Posted by Eric: Mon Jun 18th, 2012 18:20 | 14th Post |
jk wrote:Yes camera B as well for me! Agree ...camera B. D700 v D800 ?
____________________ Eric |
| Posted by richw: Mon Jun 18th, 2012 21:27 | 15th Post |
| One more for B although they are both pretty good.
|
| Posted by jmestes: Tue Jun 19th, 2012 09:30 | 16th Post |
| D3s vs D4?? Seems like B has a bit more detail.
|
| Posted by Gert: Tue Jun 19th, 2012 10:53 | 17th Post |
Don't worry JK, they are both Nikon cameras, and you probably own the same models (or better), so me posting here is not going to cost you any money
|
| Posted by Gert: Tue Jun 19th, 2012 10:55 | 18th Post |
Eric wrote: jk wrote:Erik, you are sharp as always, you got one spot on and the other is very close.
|
| Posted by Gert: Tue Jun 19th, 2012 11:10 | 19th Post |
| When most "review" sites does noise test they always shows the result at full resolution, which also makes sense if you want to see the quality of each pixel, but when you start comparing cameras with different resolution - I for one - fail to see the usefulness. Because the final output will need different scaling; e.g. if you want to print at a specific size with a given DPI, then you will have to scale the picture resolution to meet that target, and with different camera resolutions the scaling is going to be different. Up-sampling a file is going to make noise more evident, while downscaling could have the effect of "iron out" some of the noise, although it might not always be the case. So I asked myself: what if 12MP is enough, how noisy is the D800? And the result was a bit of a surprise (to me at lest), hence this post: Camera A the old D3 and Camera B the new D800, where Aftershot Pro rescaled to 12MP as the last step of the RAW->JPG conversion. I have been at true believer that more pixels equals more noise, and at the full 36MP the D800 looks more noisy than the D3, but as the pictures shows when downscaling the D800 to 12MP it is the other way around. /Gert
|
| Posted by Squarerigger: Tue Jun 19th, 2012 11:17 | 20th Post |
| I am having a very hard time picking one over the other. In most of the pics, A is sharper if I look at the switch which I assume is the auto or manual mode - has M and A on it. B is sharper if I look at the number 250 on the shutter speed dial. So, all you folks with more experience, what am I missing? D3s vs D700?
____________________ -------------------------------------------- Gary |
| Posted by Gert: Tue Jun 19th, 2012 11:42 | 21st Post |
Squarerigger wrote: I am having a very hard time picking one over the other. In most of the pics, A is sharper if I look at the switch which I assume is the auto or manual mode - has M and A on it. B is sharper if I look at the number 250 on the shutter speed dial.Even though I used liveview and tried to focus at the 125 on the shutter speed, it did not work out that way, and you are right, the A pictures are focused slightly in "front on" the B photos. But, since this was mainly meant as a noise test, you can still see a little more blue noise on the A pictures compared to the B pictures; or perhaps I should say, that in the given light the D3 has a touch more noise in the blue channel than the D800 when down-sampled to 12MP. BTW, the A/M switch decides whether the lens aperture is open or closed.
|
| Posted by Squarerigger: Tue Jun 19th, 2012 12:05 | 22nd Post |
Gert wrote:Squarerigger wrote:I am having a very hard time picking one over the other. In most of the pics, A is sharper if I look at the switch which I assume is the auto or manual mode - has M and A on it. B is sharper if I look at the number 250 on the shutter speed dial.Even though I used liveview and tried to focus at the 125 on the shutter speed, it did not work out that way, and you are right, the A pictures are focused slightly in "front on" the B photos. Thanks Gert, that explains it very well. I pick D3 vs D800 
____________________ -------------------------------------------- Gary |
| Posted by Robert: Tue Jun 19th, 2012 12:07 | 23rd Post |
Squarerigger wrote:So, all you folks with more experience, what am I missing? Look at the noise, not sharpness, the coloured blotches in the black leather area and the gritty colour on the satin chrome. Camera 'B' is much cleaner. I would be very interested to see the originals, before re-sizing the resolution.
____________________ Robert. |
| Posted by Squarerigger: Tue Jun 19th, 2012 12:42 | 24th Post |
Robert wrote:Squarerigger wrote: Thanks Robert, learning more every day.
____________________ -------------------------------------------- Gary |
| Posted by Ed Matusik: Tue Jun 19th, 2012 13:24 | 25th Post |
| I'm impressed. May have to rethink my first impulse to not buy an 800. - EdM
|
| Posted by Eric: Tue Jun 19th, 2012 13:24 | 26th Post |
Gert wrote: Eric wrote:In truth I guessed you got an 800 but I recognised the noise cut off point in camera A as either the D3 or the D700 sensor...I wrongly jumped for the D700.jk wrote:Erik,  Using the D3, as I have for some time, I have come to recognise its cut off point of 2000ISO when using artificial light. In your series it shows this very clearly ....1600 no significant noise....3200 noticeable noise. In daylight or higher kelvin lighting the noise threshold is nearer 4000ISO. For this reason I wanted to see some incandescent low kelvin lighting shots with the D800 before committing to the large files. It needs to be noise free (for me) in long exposure interior shots. At the moment if I need to use really high ISO with interior lighting, I still tripod mount and take 3 identical shots. Place the 3 shots on seperate layers in PS and drop the opacity in the 2nd and 3rd layers to 25% and 40%. This gets rid of most noise. But of course you need to have a static subject! I confess I am still trying to understand why the D800 lost noise on sampling down?
____________________ Eric |
| Posted by Gert: Tue Jun 19th, 2012 14:24 | 27th Post |
Robert wrote: Squarerigger wrote:The D3 is (obviously) at its original resolution, but here is a iso 6400 crop from the 36MP D800. If you would like to see other ISOs, let me know, and if you want the full size image, PM me. Attachment: crop-DSC_0385-6400.jpg (Downloaded 28 times)
|
| Posted by Gert: Tue Jun 19th, 2012 14:36 | 28th Post |
Eric wrote: I wouldn't say that it lost noise, it more like the noise is less evident - but i have just posted a iso 6400 crop from the full size file - tell me what you see? My personal theory for why down-sampling would "reduce" the noise goes like this: Noise is a random pattern and when downsampling from 36MP to 12MP it is like combining 3 pixels into one, now imagine one of the pixels has a wrong value and the other 2 is close to good, then we would see number 1 as noise, but when combing the 3 pixels the good values of the 2 pixels would even out the "bad" value on pixel one and the noise would thus be evident. Another way of looking at it is, when down-sampling you will lose some fine detail (12MP cannot represent the same level of detail as 36MP), and if the noise is fine-grained enough then that will be some of the "details" you loose.
|
| Posted by Robert: Tue Jun 19th, 2012 14:46 | 29th Post |
Gert wrote:The D3 is (obviously) at its original resolution, but here is a iso 6400 crop from the 36MP D800. Thank you Gert, I see now. As I thought the random pixels are much finer and being combined by the Bibble software which I believe is very good at re-sizing/re-sampling, has effectively smoothed the somewhat harsh, gritty D800 noise to something much more acceptable. I like it! Capture at high resolution and display at lower resolution... Neat. I wonder if this would also fix the awful, gritty high ISO noise from the D7000?  Doh! I just saw your second post!
____________________ Robert. |
| Posted by Gert: Tue Jun 19th, 2012 14:46 | 30th Post |
Eric wrote: Now I am curious, if the subject is static and you have a tripod, why not simply go for base iso and a long shutter time? Are you up against some very low frequency vibrations?
|
| Posted by Gert: Tue Jun 19th, 2012 14:52 | 31st Post |
Robert wrote: Good point Robert, the result will vary depending on the software used.
|
| Posted by Robert: Tue Jun 19th, 2012 15:49 | 32nd Post |
Gert wrote:Robert wrote:Good point Robert, the result will vary depending on the software used. Not just that, on some software there are various ways to re-sample, it might be worth experimenting? I think in a way this is a variation on Eric's method.
____________________ Robert. |
| Posted by Robert: Tue Jun 19th, 2012 16:34 | 33rd Post |
| Hope you don't mind but I have just run that full resolution crop through five of the standard Photoshop Cs5 image reduction processes reduced from 602 to 200 pixels high. Very interesting... There is quite a difference between the levels of noise. I think the forum software is reducing it again but they still seem to be very similar to what I saw before I uploaded. My money is on the top image the default, Bi-Cubic. Attachment: Group of 5.jpg (Downloaded 25 times)
____________________ Robert. |
| Posted by jk: Tue Jun 19th, 2012 17:17 | 34th Post |
| Well I'm happy with my D800 and D3 and D3S so the result doesnt worry me. The results A/B were close by the B ones did seem better to me.
____________________ Still learning after all these years! https://nikondslr.uk/gallery_view.php?user=2&folderid=none |
| Posted by richw: Tue Jun 19th, 2012 23:12 | 35th Post |
| That is very, very, interesting thanks for posting. In Eric's method (Cambridge in color method?) you are averaging out three pixels by laying them on top of one another. By downsampling you are effectively doing the same thing taking three pixels in a group and averaging them (I know the reality is more complicated but this seems roughly what is happening). So this might make Eric tripod less????
|
| Posted by Robert: Wed Jun 20th, 2012 02:21 | 36th Post |
| With the fine spacing of the pixels involved with a D800 or D7000 I doubt Eric is hitting the exact same spot with each exposure (pixel for pixel), even bolted down to a solid tripod and using a remote.
____________________ Robert. |
| Posted by richw: Wed Jun 20th, 2012 03:30 | 37th Post |
Robert wrote:With the fine spacing of the pixels involved with a D800 or D7000 I doubt Eric is hitting the exact same spot with each exposure (pixel for pixel), even bolted down to a solid tripod and using a remote. True, but the general principle is still true, and too much variation would be an unsharp shot.
|
| Posted by Eric: Wed Jun 20th, 2012 07:35 | 38th Post |
Gert wrote: Eric wrote:Its very useful rethinking why we do things from time to time. You have made me do that.Now I am curious, if the subject is static and you have a tripod, why not simply go for base iso and a long shutter time? I started the 3 layer averaging (for some work) some years ago when I changed from the D1X to the D2X... because I found two things 'wrong' with it. The extra pixels (at that time ) brought more noise. But I also felt the extra pixels required greater precision in technique. Simply ...I had to half my shutter speeds to get the same IQ as the D1X. At that time I was shooting old hotel interiors and I found people walking about in adjacent rooms bounced the floor!! So I used a tripod but tried to to keep more normal (hand held) shutter speeds. I also used flash, to avoid some mixed lighting WB issues, so getting nearer the sync speed was desireable also. This meant with the high ISO noise of the D2X I had to do the Cambridge in Colour trick. But things change and maybe this D800 feature you have discovered has changed the ground rules? Very interesting.
____________________ Eric |
| This is topic ID = 224 | ||
| Nikon DSLR Forums > Camera and Lens Forums > Cameras > High iso noise "blind test" | Top | |
Users viewing this topic |
||
Current theme is Blue
| A small amount of member data is captured and held in an attempt to reduce spammers and to manage users. This site also uses cookies to ensure ease of use. In order to comply with new DPR regulations you are required to agree/disagree with this process. If you do not agree then please email the Admins using info@nikondsl.uk Thank you. |
Hosted by Octarine Services
UltraBB 1.173 Copyright © 2008-2024 Data 1 Systems
Page processed in 0.1262 seconds (71% database + 29% PHP). 244 queries executed.
UltraBB 1.173 Copyright © 2008-2024 Data 1 Systems
Page processed in 0.1262 seconds (71% database + 29% PHP). 244 queries executed.