fredmiranda.com Login
This topic is in the 2010 archive. You can read it and like its posts here. It is read only.

Archive 2010Noise, Interpolation, and Comparison 101

  
 
Breitling65
Offline
Details
p.3 #1Noise, Interpolation, and Comparison 101
mk IV for sure king of "NOT IN STOCK" ...
Feb 03, 2010 at 04:21 AM
Pixel Perfect
Offline
Details
p.3 #2Noise, Interpolation, and Comparison 101
BigBlueDodge wrote:
I think one obvious point we are all missing here is to appreciate what a damn fine camera the 5D classic was and still is. It's impressive that almost 5 yr old technology is still hanging in there with the new kids on the block.


Big pixels in a big sensor = excellent


Hey I'd still take a 1D IV over a 5D any day, but the old 5D is still superb in IQ.
Feb 03, 2010 at 04:35 AM
Beni
Offline
Details
p.3 #3Noise, Interpolation, and Comparison 101
Given that the 5D's iso 3200 is not even native and it's 5 years older, methinks the result is not that impressive for the mk4.

5 years after the 5D, the mk4 is not as good at what it is advertised to do (18 megapixels) as the 5D was at doing what it wasn't even advertised to do (13 megapixels at iso 3200, you had to enable a CF just to get it to work).
Feb 03, 2010 at 04:38 AM
Pixel Perfect
Offline
Details
p.3 #4Noise, Interpolation, and Comparison 101
Hey, I've seen tests showing much better results for the 1D IV. I wouldn't be putting too much value on one test like this and I'm not having a go at the OP at all. Every time a new camera comes out it's the same thing over and over.
Feb 03, 2010 at 04:47 AM
thw2
Offline
Details
p.3 #5Noise, Interpolation, and Comparison 101
Have you guys seen this which shows RAW files comparison at the PIXEL level between 1D4, 1Ds3, 7D and D3s?

Outcome:
1D4 is 1 stop better than 1Ds3 (which is equivalent to 5D2).
1D4 is 1.5 stops better than 7D.
1Ds3 is merely 0.5 stop better than 7D.
D3s is 1 stop better than 1D4.

This means D3s is TWO whole stops better than 1Ds3.

OK... better duck for cover... I can see all the guns pointing at me already..
Feb 03, 2010 at 05:40 AM
Pixel Perfect
Offline
Details
p.3 #6Noise, Interpolation, and Comparison 101
juza doesn't like the APS-H format, he's got it in for the 1D IV from the start

D3s looks amazing for high ISO work.
Feb 03, 2010 at 06:43 AM
Dawei Ye
Offline
Details
p.3 #7Noise, Interpolation, and Comparison 101
32067dlm wrote:
Okay, I did another test with a little more control. This time I ensured closer exposure. I upsized the 5D file to match the 1DIV file using Bicubic Smoother. How many more ways shall I do this?

stanj wrote:
Don't. You'll never achieve consensus. What I think, or what others think doesn't really matter, it will always be a mud bath because people will always measure differently and stubbornly believe their approach is the only correct one. I know which approach is correct for me, but that doesn't matter. Nor do I personally think that this comparison matters, because few people will be deciding between the 1D4 and the 5D1 based on high ISO noise alone, or even predominantly. YMMV of course.


+1

brainiac wrote:
NO! DO NOT COMPARE NOISE PER UNIT AREA! Compare noise per image.

Ariel Bravy wrote:
It depends on what you're looking for and how you shoot.

IMHO, both are valid comparisons.

The other day I compared my 4MP 1D vs. a friend's 21MP 5DII. I wanted to see how the high ISO compared. Did I uprez the 4MP to 21MP to do a fair comparison? No. Why not? Because it has nothing to do with my reality and the way I like to take pictures and look at images. I never do that in real life so at that point we're talking technicalities and not practical shooting. It's really helpful to know as part of the
...Show more →

+1

Nowadays I'm becoming a bit disturbed by the fanaticism that many posters have (and I'm not talking about anyone in particular - just in general) in the sense that anytime someone posts a 100% crop there comes a hail of posts claiming "NO THIS TEST SUX CAUSE YOU DIDN'T UPSIZE/DOWNSIZE/EQUALIZE!!!!!!!1111"

It is impossible to deny that equalising resolution makes a lot of logical and technical sense for comparison purposes - but at the same time, I think posters in that camp also need to appreciate it is not the only metric relevant to many photographers.

I believe looking at pixel level noise is also very important, because the pixel level is what we can basically push the file to. If we purchase a high res camera expecting to use that higher resolution, then it is also important to understand the boundaries of performance.

The advantages of a high pixel density camera is often reach...that advantage is nullified if one has to downsize photos to a large pixel camera like a 5D in order to maintain same high ISO performance as past generation cameras


Edited on Feb 03, 2010 at 07:17 AM · View previous versions
Feb 03, 2010 at 07:16 AM
mfurman
Offline
Details
p.3 #8Noise, Interpolation, and Comparison 101
Ariel Bravy wrote:
brainiac wroteNO! DO NOT COMPARE NOISE PER UNIT AREA! Compare noise per image.
It depends on what you're looking for and how you shoot.

IMHO, both are valid comparisons.

The other day I compared my 4MP 1D vs. a friend's 21MP 5DII. I wanted to see how the high ISO compared. Did I uprez the 4MP to 21MP to do a fair comparison? No. Why not? Because it has nothing to do with my reality and the way I like to take pictures and look at images. I never do that in real life so at that point we're talking technicalities and not practical shooting. It's really helpful to know as part of the
...Show more →

I completely agree! My thoughts exactly.
Feb 03, 2010 at 07:16 AM
Dawei Ye
Offline
Details
p.3 #9Noise, Interpolation, and Comparison 101
thw2 wrote:
Have you guys seen this which shows RAW files comparison at the PIXEL level between 1D4, 1Ds3, 7D and D3s?

Outcome:
1D4 is 1 stop better than 1Ds3 (which is equivalent to 5D2).
1D4 is 1.5 stops better than 7D.
1Ds3 is merely 0.5 stop better than 7D.
D3s is 1 stop better than 1D4.

This means D3s is TWO whole stops better than 1Ds3.

OK... better duck for cover... I can see all the guns pointing at me already..


The problem with these "1 stop better than..." interpretations is that they never, ever make sense, nor are consistent.

On the same page, Juza also says:
I did not have a 1D Mark3 in my hands during this test, but I tried it in the past and it was about 1 stop better than 1DsIII

So according to your info, that's saying the 1D Mark 3 is as good as the 1D4 and 1 stop better than the 5D Mark II, which many tests (e.g. the-digital-picture.com) have shown is better than the 1D4 and 1Ds3 and 1D3.

i.e.
1D3 > 1Ds3
1D4 > 1Ds3
Therefore,
1D3 = 1D4
1D3 > 1Ds3
1Ds3 = 5D2
1D3 > 5D2
5D2 > 1D4
Since 1D4 = 1D3,
Therefore, 5D2 > 1D3
yet 1D3 > 5D2



Can you see all of these findings end up seeming like total contradictory garbage? We never seem to be able to reach consensus

Don't get me started about Nikon...their on chip noise reduction is so strong there's no point comparing with Canon cameras
Feb 03, 2010 at 07:22 AM
alundeb
Offline
Details
p.3 #10Noise, Interpolation, and Comparison 101
One often overlooked problem with viewing pixels at 100%, is the visual effect of pixel boundaries. These square boxes with abrupt changes in intensity between them, affect how we percept the image. To avoid that, one could interpolate to 200% or higher, to make all images look soft at pixel level. Then the quality of the images can be compared, with the disturbing element of pixel boundaries relatively diminished and moved away from the sharpness limit of the image content. It is not a practically good solution, becuase it is a waste of bandwidth and montior resolution.

I thought I would just mention this crazy idea, otherwise I agree with many of the comments about different comparison methods for different purposes.
Feb 03, 2010 at 07:37 AM
David Manning
Offline
Details
p.3 #11Noise, Interpolation, and Comparison 101
Pixel Perfect wrote:
Hey I'd still take a 1D IV over a 5D any day, but the old 5D is still superb in IQ.


I'm very happy to have both.
Feb 03, 2010 at 08:05 AM
kakomu
Offline
Details
p.3 #12Noise, Interpolation, and Comparison 101
brainiac wrote:
32067dlm wroteThis is a 100% comparison...
So it's not a comparison. To compare noise you need to equalise magnification. That means uprezzing the 5D file to the same pixel dimensions as the 1D4 file before comparing crops at 100%.


uprezzing magnifies noise and is apart from what people typically do with images (If people PRINT images at large sizes, that's more akin to an analog upres, not digital), which gives any higher resolution sensor an instant edge. He would need to downres the 1D image and compare.
Feb 03, 2010 at 08:09 AM
leewoolery
Offline
Details
p.3 #13Noise, Interpolation, and Comparison 101
sivrajbm wrote:
Canon ISO yes, Nikon D3s is the KING of camera ISO...
As my grandson says Sorry bout that...


I agree...I've shot sports with the 1D Mark IV and D3s and was very disappointed with the high ISO performance of the Mark IV. Even the D700 and original D3 have better noise at ISO's from 1600 to 10,000...but the D3s is the king.

For studio portrait work at ISO 100, the images from the Mark IV are stunning.

For strobed or daylight, low ISO action photography, the Mark IV will be hard to beat.

Much success,

Lee Woolery
Speedshot
Feb 03, 2010 at 08:35 AM
David Manning
Offline
Details
p.3 #14Noise, Interpolation, and Comparison 101
kakomu wrote:
uprezzing magnifies noise and is apart from what people typically do with images (If people PRINT images at large sizes, that's more akin to an analog upres, not digital), which gives any higher resolution sensor an instant edge. He would need to downres the 1D image and compare.


In this thread alone, I have done three comparisons. I compared the two at the pixel level (and got flamed), I down sized the 1D4 file, and lastly, I upsized the 5D file. In each case, there appears to be more noise in the 1D4 files. I do agree with your logic. I think it contaminates the test to uprez a smaller file, and if I change image size, 99% of the time, it's to make the file smaller for a smaller print.

David
Feb 03, 2010 at 09:01 AM
SLD
Offline
Details
p.3 #15Noise, Interpolation, and Comparison 101
leewoolery wrote:
For studio portrait work at ISO 100, the images from the Mark IV are stunning.

For strobed or daylight, low ISO action photography, the Mark IV will be hard to beat.

Speedshot


Does this mean the Mark IV has better IQ then Nikon D3 and D3s under ISO 1600?

Thanks
Feb 03, 2010 at 09:01 AM
kakomu
Offline
Details
p.3 #16Noise, Interpolation, and Comparison 101
32067dlm wrote:
got flamed


That's the Canon Mount Forum for you.
Feb 03, 2010 at 09:33 AM
leewoolery
Offline
Details
p.3 #17Noise, Interpolation, and Comparison 101
lidesun wrote:
Does this mean the Mark IV has better IQ then Nikon D3 and D3s under ISO 1600?

Thanks


I've always found the noise levels and overall image quality from the Mark III at ISO 100-200 to be excellent and the Mark IV has even better performance...plus a bigger file.

From my personal tests...for anything above ISO 400, I don't think any camera beats the D3s with regards to noise.

I'd encourage anyone to rent the cameras and do your tests before making a purchase.

I was really hoping for the high ISO performance from the Mark IV to be equal to or better than the D3s...but it's not even close.

There's more to a camera than how it performs at ISO 800 and above but for my business, it's vitally important to have the lowest possible noise at the highest ISO levels.

Much success,

Lee Woolery
Speedshot

Feb 03, 2010 at 10:28 AM
brainiac
Offline
[X]
Details
p.3 #18Noise, Interpolation, and Comparison 101
kakomu wrote:
brainiac wrote
32067dlm wroteThis is a 100% comparison...
So it's not a comparison. To compare noise you need to equalise magnification. That means uprezzing the 5D file to the same pixel dimensions as the 1D4 file before comparing crops at 100%.
uprezzing magnifies noise and is apart from what people typically do with images...


Uprezzing is precisely what your printer does before it puts all those ink drops on the paper. It is drivel to suggest that uprezzing makes comparison between 18 Mpixel and 13 Mpixel cameras unfair. Uprezzing using bicubic (not a variant) produces an image which will be indistinguishable in print from the original, UNLESS the print is so large that the boundaries between pixels are resolved.

The reason why uprezzing is a fairer method of comparison than downrezzing is that downrezzing has the potential to destroy detail, and therefore can hide an advantage that the higher rez file may have over the uprezzed file. For comparison of noise alone, downrezzing the higher rez file is perfectly adequate EXCEPT that noise reduction algorithms are less destructive when performed on the full rez file than on the downrezzed file. In tests like the ones shewn on this thread, the iso is such that noise reduction would very likely be used in real world output, so any comparison of noise reduced output should have the noise reduction applied BEFORE downrezzing.
Feb 03, 2010 at 11:09 AM
brainiac
Offline
[X]
Details
p.3 #19Noise, Interpolation, and Comparison 101
32067dlm wrote:
In this thread alone, I have done three comparisons. I compared the two at the pixel level (and got flamed), I down sized the 1D4 file, and lastly, I upsized the 5D file. In each case, there appears to be more noise in the 1D4 files.


Here they are:
http://www.davidmanningphotography.com/Mark-IV/MarkIVNoise.jpg
http://www.davidmanningphotography.com/Mark-IV/Noise_Compare_2.jpg
http://www.davidmanningphotography.com/Mark-IV/Noise_Compare.jpg

It is clear to me that the first one, where equal magnification was not ensured, tells a different story to the other two. In your first sample the 1D4 looks significantly worse than the 5D, whereas in the subsequent correctly made comparisons the shortfall is much less significant, to the point where I would say that after good noise reduction the 1D4 will probably match the larger-sensor of the 5D for noise, and probably beat it for detail. As someone has already mentioned, the 5D may lose its edge in lower light because it suffers from banding in shadows that we can not see here.

> I think it contaminates the test to uprez a smaller file, and if I change image size, 99% of the time, it's to make the file smaller for a smaller print.

Do you have any reason to suppose this 'contamination', or are you just repeating folklore?

I can only repeat the scientific truth that is self-evident to those who understand this problem properly: comparison "at the pixel level" between a 16 Mpixel camera and a 13 Mpixel camera is largely futile because it doesn't give you accurate data about how the cameras perform when their frames are fully used. The crops above illustrate this point perfectly as they show that magnification seriously affects our impressions of camera performance.
Feb 03, 2010 at 11:15 AM
brainiac
Offline
[X]
Details
p.3 #20Noise, Interpolation, and Comparison 101
Here are the results with a typical dose of chroma NR applied. Interestingly, in the crop showing the native 1D4 against the uprezzed 5D, the 1D4 holds its own better than in the other crops. That may be because in that case the 1D4 is not being penalised by incorrect enlargement, and noise reduction is being applied in a more fine-grained way.

http://cyberphotographer.com/1d4/1d4crops.jpg

Don't get me wrong, I'm not saying that the 1D4 is better or worse than the 5D, but if we're going to talk about that question it would be a good start to make the tests reflect what people will see in actual real world prints from the two cameras, instead of what they will see when zooming into 100% to make bogus comparisons at differing enlargement ratios.

Edited on Feb 03, 2010 at 12:07 PM · View previous versions
Feb 03, 2010 at 11:38 AM
FM ForumsCanon ForumJoin Upload & Sell
  
 

You are not logged inLog inRegister