Go Back   Steve's Digicams Forums > Digicam Help > Newbie Help

Reply
 
Thread Tools Search this Thread
Old Feb 12, 2010, 2:48 PM   #21
Senior Member
 
rfortson's Avatar
 
Join Date: May 2006
Posts: 349
Default

Quote:
Originally Posted by peripatetic View Post
There is another advantage. When I first started out on DSLR 6 years ago I became convinced that shooting RAW was a good idea. I'm very glad I did, as the RAW converters of today are much better than they were back then. I can now revisit shots I took 6 years ago and get a better image from them than I did back in 2004.
That is definitely true, and not often touted as a benefit to shoot raw. While jpeg processors built into cameras continually improve, when you shoot jpeg, that's as good a shot as you'll ever get from that file. When you shoot raw, you can take advantage of the improvement in raw processors (and computers) for as long as that file is still readable.
rfortson is offline   Reply With Quote
Old Feb 12, 2010, 5:45 PM   #22
Senior Member
 
Ordo's Avatar
 
Join Date: Jul 2009
Location: BsAs
Posts: 3,452
Default

Raw vs jpeg seems to me as another market compromise we should asume. For good or for bad.
Ordo is offline   Reply With Quote
Old Feb 12, 2010, 6:23 PM   #23
Administrator
 
Join Date: Jun 2003
Location: Savannah, GA (USA)
Posts: 22,378
Default

One of the problems I'm seeing with some camera models (including models from Sony, Pentax, Nikon and others) is that there appears to be some level of Noise Reduction at the raw level. For example, some of the newer Sony models have this characteristic, and evidence *very* strongly suggests the same thing is happening with some Nikon models (like the D90, as well as some of the Pentax dSLR models. Ditto for some Olympus models.

So, because manufacturers are trying a bit *too hard* to look good on popular tests showing performance at the raw level, it limits a user's ability to extract the most out of a camera's raw images later (because the sensor data is being manipulated at the raw level).

Unfortunately, a lot of the current tests of a camera's ability based on raw files are usually taken as "gospel" on how one sensor compares to another, even though the [supposedly expert] testers may not have [yet] realized the manipulation of raw data going on by some of the camera manufacturers (as it tends to be a bit subtle and hard to detect unless you know what to look for using more sophisticated tools for raw file analysis).

IOW, I'd take any tests of Signal to Noise ratios and more with a "grain of salt", as those tests can be fooled by raw file manipulation. Instead, let your eyes be the best judge of how images from different cameras compare to each other.
JimC is offline   Reply With Quote
Old Feb 13, 2010, 6:42 AM   #24
Super Moderator
 
peripatetic's Avatar
 
Join Date: Nov 2004
Posts: 3,599
Default

Quote:
IOW, I'd take any tests of Signal to Noise ratios and more with a "grain of salt", as those tests can be fooled by raw file manipulation. Instead, let your eyes be the best judge of how images from different cameras compare to each other.
Of course in one respect that's all that matters - it's your cash, so buy what looks good to you.

But you radically overestimate the ability of any person's eye to make meaningful and reliable distinctions in such matters. DXOMark may be imperfect and/or controversial, but it's repeatable and the methodology is published for anyone to replicate. In other words, it's science and based on hard data, I'll take that over anecdotes any day.
__________________
My gallery
My X100 blog
peripatetic is offline   Reply With Quote
Old Feb 13, 2010, 7:03 AM   #25
Administrator
 
Join Date: Jun 2003
Location: Savannah, GA (USA)
Posts: 22,378
Default

I don't ignore their tests. But, I would sure supplement them using my own eyes as a guide.

For example, if you look at Sensor numbers, Signal to Noise Ratio comparisons and more between the Nikon D300 and D90, you'd assume that the D90 has much better image quality.

But, if you look at the photos, you may come to a totally different conclusion. For example, if you look at the D90 review at dpreview.com, they have some comparisons between these cameras at both the raw and jpeg level. They indicated that no amount of reshooting could get the D90 to match the D300's output (and they made similar comments about them for both raw and jpeg images).

It's also very difficult to judge the quality of things like noise patterns by looking at their graphs. Originally, they had .pdf files online for cameras tested that allowed you to see the patches they were measuring to get a better idea of how cameras compared. But, AFAIK, they don't let you see that anymore. Instead, you're stuck looking at numbers and graphs to try and tell how image quality compares (which IMO, by themselves, can be very misleading).
JimC is offline   Reply With Quote
Old Feb 13, 2010, 7:15 AM   #26
Super Moderator
 
peripatetic's Avatar
 
Join Date: Nov 2004
Posts: 3,599
Default

Quote:
For example, if you look at Sensor numbers, Signal to Noise Ratio comparisons and more between the Nikon D300 and D90, you'd assume that the D90 has much better image quality.
Then you really need to read the charts and the site info on interpretation again, because that is absolutely not what they suggest IMO; it suggests that the sensor performance is almost identical.


Quote:
It's also very difficult to judge the quality of things like noise patterns by looking at their graphs. Originally, they had .pdf files online for cameras tested that allowed you to see the patches they were measuring to get a better idea of how cameras compared. But, AFAIK, they don't let you see that anymore. Instead, you're stuck looking at numbers and graphs to try and tell how image quality compares (which IMO, by themselves, can be very misleading).
The image is integrated in the page now. Hover your mouse over the slider bars at the right of the graph and it shows the underlying image used to make the measurements.

I believe it is always sensible to normalise for the Print view because this gives the best comparison unless you are going to be printing at substantially higher resolutions than one of the two cameras allows.

http://dxomark.com/index.php/eng/Image-Quality-Database/Compare-cameras/%28appareil1%29/294|0/%28appareil2%29/295|0/%28onglet%29/0/%28brand%29/Nikon/%28brand2%29/Nikon

Comparing the D90 to the D300 shows about a half stop advantage to the D90 overall, but the differences are very small and chiefly in DR and color and tone sensitivity in low-light conditions.

Quote:
But, if you look at the photos, you may come to a totally different conclusion.
Actually being able to judge margins than fine by eye? Good luck. You may be able to convince yourself that you can tell them apart, but I'd be willing to bet under controlled double-blind conditions you wouldn't have a snowball in the Sahara's chance. :-)

aka - the cameras sensors are essentially equivalent for general purpose photography. Though of course the D300 is a much better camera in other respects.
__________________
My gallery
My X100 blog

Last edited by peripatetic; Feb 13, 2010 at 7:29 AM.
peripatetic is offline   Reply With Quote
Old Feb 13, 2010, 7:28 AM   #27
Administrator
 
Join Date: Jun 2003
Location: Savannah, GA (USA)
Posts: 22,378
Default

Interesting. But, I don't think those are the actual images used for the measurements (rather they appear to be designed to give you an idea of what to expect at a given point on the graph).

The .pdf files they used to allow you to download with test results, included the actual gray and color patches being used for their measurements.
JimC is offline   Reply With Quote
Old Feb 13, 2010, 7:41 AM   #28
Administrator
 
Join Date: Jun 2003
Location: Savannah, GA (USA)
Posts: 22,378
Default

Quote:
Originally Posted by peripatetic View Post
Actually being able to judge margins than fine by eye? Good luck. You may be able to convince yourself that you can tell them apart, but I'd be willing to bet under controlled double-blind conditions you wouldn't have a snowball in the Sahara's chance. :-)
Some experienced reviewers may disagree with you (for example, Phil Askey and crew). ;-)

It's pretty easy for me to see that D300 images look better when viewing some of the comparisons taken in controlled conditions.

Things like S/N ratio tests can be very misleading in the amount of detail being retained. You'll even see articles at the Dx0Mark site discussing how NR at the raw level impacts their test results (which is why they're using extrapolated data for some cameras models now, where they've been able to detect manipulation of data at the raw level).

However, some pretty smart people have presented very strong evidence of NR at the raw level for cameras that they have not yet detected doing it, using pretty sophisticated raw analysis tools to reach their conclusions. They're looking for more obvious smoothing of data and there are more subtle techniques that make manipulation of data at the raw level much harder to detect without using a different analysis approach. You can end up with softer images and still have a great S/N ratio on most tests.
JimC is offline   Reply With Quote
Old Feb 13, 2010, 7:57 AM   #29
Administrator
 
Join Date: Jun 2003
Location: Savannah, GA (USA)
Posts: 22,378
Default

Now, some tests are more sophisticated. For example, if you look at Imatest results, you can see a bit more about a camera's noise pattern (versus just S/N ratios). You'll see a lot of discussion in Dave Etchells' reviews when looking at Imatest results. For example, here are the results for the D90:

http://www.imaging-resource.com/PROD...D90IMATEST.HTM
JimC is offline   Reply With Quote
Old Feb 13, 2010, 8:18 AM   #30
Administrator
 
Join Date: Jun 2003
Location: Savannah, GA (USA)
Posts: 22,378
Default

P.S.

Here's a discussion of D90 raw file analysis showing evidence of high ISO NR that you may find interesting:

http://forums.dpreview.com/forums/re...ssage=32401883

Nikon also appears to be performing long exposure NR at the raw level with some models, even with long exposure NR turned off. Here's an article discussing it:

http://www.astrosurf.com/buil/nikon_test/test.htm

They also appear to do other types of preprocessing of raw data. For example, Dave Coffin (the author of dcraw.c ) noticed that Nikon was modifying raw data years ago with some of their cameras (applying multipliers to raw data). See question #12 from this interview with Dave in 2005:

http://www.dpreview.com/news/0504/05...ninterview.asp

In fact, I had a long conversion with Dave the night before that interview (we were mostly discussing Nikon's encryption of RGB multipliers for White Balance at the time, and he had offered to do an interview with Steve's at the time, too). We've also had a number of discussions since then (mostly about dcraw.c, it's licensing, etc.).
JimC is offline   Reply With Quote
 
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 2:00 AM.