Jump to content

Canon EOS 400d XTi noise tests posted at dpreview


jonas_gustavsson

Recommended Posts

Finally, fellow pixelpeepers, dpreview has posted some data on the 400D noise

levels as part of a comparison with Nikon D80 in a recent review: <a

href="http://www.dpreview.com/reviews/nikond80/page18.asp">400d et al

noise</a> (also data on other comparison pages)<p>

 

Still waiting for the real stuff, of course, but maybe this is an indication

that the noise doesn't quite render the increase 8->10MP useless, but that we

are very much approaching the limit of what is possible. Since I'm used to

ISO800 noise on a D60, I still think they should have treated us to a hardware

3200 level on the 400d for emergencies.<p>

Link to comment
Share on other sites

You will notice that the Nikon and Sony show LESS DETAIL then the XTi frame for frame.

 

Nikon and Sony have noisier sensors, so they have to apply more aggressive noise reduction algorithms, which does tame the noise some, but at the EXPENSE of image detail.

 

The fact that Nikon and Sony are able to nearly match the noise performance of the Canon seems great to the untamed eye, but look closely....detail is smeared out, and if you see this then you'll not get suckered into thinking that the Nikon & Sony are noise-competitive with the Canon XTi.

 

It's funny how many Nikon users discount the noisy images coming from even their top-of-the-line bodies by saying "...with NoiseNinja applied the noise is removed....", but what they don't say is that image detail is also removed.

 

But all this aside, Phil's noise comparos still show the Canon doing better.

 

 

Canon is still King of Noise-Performance.

Link to comment
Share on other sites

I agree that the patches paint a rosier picture for Canon than the graphs. It would be neat to see some tests where the potential excessive smearing of post-processing has been corrected for, though. Maybe some form of deconvolution applied to "re-sharpen" an in-focus edge in a shot could be used for this.

 

Personally, I think it is sad that the Sony does not fare better - I want this to be a tough competetive field rather than a duopoly.

Link to comment
Share on other sites

Until I see the Noise of these cameras against the current benchmark -> the 30D/20D -> I think the jury is still out regarding noise. There is not yet any reason to believe that the 400XT is any worse than that benchmark.

 

In other words. . there is no real reason to believe that any limit has been reached.

Link to comment
Share on other sites

Here's another way of looking at this noise business...theoritically Sony could make it's 1600 ISO less noisy then the current noise-performance king, the Canon 5D...sure it can do this easily by making the anti-noise algo's it uses way more aggressive, and then it could have bragging rights in all the industry mags, and in fact most of the masses of consumers will see this and think "wow Sony has the least noisy cameras".

 

But the bad news is that image detail will be smeared away beyond recovery, human skin tones will be plasticy smooth, unnatural, etc.

 

Photo mags like PopularPhotography and others will not consider this, even as they say the Sony has "great noise performance"...they're not the only offenders...almost all the mags do this...it's about advertising dollars.

Link to comment
Share on other sites

Well, there is the <a href="http://www.cameralabs.com/reviews/Canon400D/page4ca.shtml">Camera Labs</a> 'real life' comparison to the 350XT, which seems to suggest pretty similar performance, but as I understand it, the 20D/30D still has a <a href="http://www.dpreview.com/reviews/canoneos350d/page19.asp">slight edge</a> in high ISO noise performance. Let's hope some Canon engineer is busy figuring out how to improve 40D further without resorting to image intensifiers of liquid nitrogen cooling. :-)<p>

Regarding smearing, I doubt that it really results in a complete loss of information, but rather in some kind of degradation. If an image is softened using gaussian blur (lowering noise RMS), unsharp mask can - to some extent - be used to recover the original sharpness. Since whenever working with a DSLR image - particularly RAW - some sharpening seems necessary, it would be very interesting to see what the output would look like from a noise perspective once the same perceived sharpness has been recovered.<p>

Compariung different resolution cameras, the inherent benefit of having more pixels to start with, allowing some averaging to bring 10 to 8 MP may also be interesting to consider to see if the 2 MP extra are pretty much "empty" or not. <p>

Link to comment
Share on other sites

Jonas, I'll have to disagree with you...once you apply Guassian blur to an image, the original details are lost forever...no amount of USM, no amount of fancy photoshopping will bring that detail back. It is gone forever.

 

When an anti-noise algorithm is applied, it "smears" pixels, which is very damaging to edges, the pixels that define the edges get blended to "remove" the noise, colors get spread. Since USM relies on those edges to create lines of contrast (threshold value) the width of whatever radius one specifies, well, if those edges are gone, or minimized, USM is not going to be very effective.

 

Some people also believe the other myth that blown highlights can be recovered....nope on that too....impossible...once a pixel's value reaches 255 (8-bit) or 4,096 (12 bit), that is it...blown forever...never to be brought back to the actual intensity value. Same for blocked up shadows.

 

This is why cameras that use too aggressive anti-noise algorithms remove FOREVER the ability to get a very sharp image.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...