nikon D300 approach to noise v canon

[woof75]

Sorry to have to do this but I really am interested, after looking at a review of the new D300 it seems that Nikon are taking a different approach to noise compared to Canon. I've never liked the soft, almost smeared images canons have produced and it seems that Nikon have produced a camera with more "grain" and less smearing. Looks very promising, anyone have experience with the two? (and be nice, lets not try and have a brand war, there only cameras).

[Graeme Nattress]

The visual look can often be down to the raw conversion demosaicing algorithms, rather than the camera itself though.

To really compare the noise, you'd have to have access to the raw data before demosaicing, and compare how the red, green and blue components of the image look with in camera NR on and off.

The thing is, if you're recording uncompressed raw, any NR you do in camera, can also  be done equally well (if you know what you're doing) in software. To my mind, it will be the raw software that, as it improves, that will improve the noise characteristics of all raw shooting cameras.

Graeme

[woof75]

The visual look can often be down to the raw conversion demosaicing algorithms, rather than the camera itself though.

To really compare the noise, you'd have to have access to the raw data before demosaicing, and compare how the red, green and blue components of the image look with in camera NR on and off.

The thing is, if you're recording uncompressed raw, any NR you do in camera, can also  be done equally well (if you know what you're doing) in software. To my mind, it will be the raw software that, as it improves, that will improve the noise characteristics of all raw shooting cameras.

Graeme
[a href=\"index.php?act=findpost&pid=181097\"][{POST_SNAPBACK}][/a]

I process in adobe camera raw and when processed through this software the nikon and camera files have different noise characteristics. It's this that I am interested in.

[DonWeston]

I recently just sold my D300 and tested the D3 and ended up with getting a 5D again. I found the D300 to soften the images too much for my tastes, with much reduced detail compared to my 40D. We are talking about only image quality, NOT body capabilities here. The D3 was in a "nother" league compared to the D300, I would guess for the full frame advantage. BUT overall I just prefered the 5D image quality, JMHO. YMMV....And it was very light on my wallet at todays prices, conjecture about a 5D mark II is just poppycock at this point, since no cameras in out for the general public to try out. Who knows what this will turn out as when they are in the public domain..

[Graeme Nattress]

I process in adobe camera raw and when processed through this software the nikon and camera files have different noise characteristics. It's this that I am interested in.
[a href=\"index.php?act=findpost&pid=181103\"][{POST_SNAPBACK}][/a]

And if Adobe put some advanced NR in there, you may no longer see a difference. Who knows? Without extensive analysis of the raw data, it's hard to tell, but we do know Nikon put NR on the raw files before recorded, and it looks like, to me, very good chroma noise reduction, but you can do that  in software on untouched RAW files too....

Graeme

[Panopeeper]

Sorry to have to do this but I really am interested, after looking at a review of the new D300 it seems that Nikon are taking a different approach to noise compared to Canon

There is nothing new here. Canon leaves the "black current" in the raw data and it's the task of the raw processors to remove it. Nikon records the "amended" data.

It is important to understand, that

1. ACR performs some small noise reduction even if NR is turned down. This is new in version 4; after lots of complains that ACR removes fine details, this inherent noise reduction has been reduced in 4.3 (or in 4.2), but not fully eliminated. ("Inherent", for this NR is part of the demosaicing.);

2. ACR's processing of the black current information is suboptimal, to say the least; the customary horizontal (in landscape orientation) streaks in dark images are the product of ACR's approach. DPP makes it better.

So, your choice of raw processor makes a greater difference for the Canon shots than for Nikons.

I've never liked the soft, almost smeared images canons have produced and it seems that Nikon have produced a camera with more "grain" and less smearing

Did you make comparable shots? Do you mind uploading them?

I suggest an extra test. I am just evaluating such shots from Canon 350D, 20D, 40D and probably more.

1. Cover the lens very tightly with the cap, or use the body cap.

2. Adjust the shutter to 5sec (5sec, so that it be comparable with the other shots I am collecting).

3. Cover the viewfinder tightly and shoot.

4. Upload the raw files.

You may want to do the same in room temperature and with a cold camera. I did not find the difference with the 40D relevant.

[woof75]

There is nothing new here. Canon leaves the "black current" in the raw data and it's the task of the raw processors to remove it. Nikon records the "amended" data.

It is important to understand, that

1. ACR performs some small noise reduction even if NR is turned down. This is new in version 4; after lots of complains that ACR removes fine details, this inherent noise reduction has been reduced in 4.3 (or in 4.2), but not fully eliminated. ("Inherent", for this NR is part of the demosaicing.);

2. ACR's processing of the black current information is suboptimal, to say the least; the customary horizontal (in landscape orientation) streaks in dark images are the product of ACR's approach. DPP makes it better.

So, your choice of raw processor makes a greater difference for the Canon shots than for Nikons.
Did you make comparable shots? Do you mind uploading them?

I suggest an extra test. I am just evaluating such shots from Canon 350D, 20D, 40D and probably more.

1. Cover the lens very tightly with the cap, or use the body cap.

2. Adjust the shutter to 5sec (5sec, so that it be comparable with the other shots I am collecting).

3. Cover the viewfinder tightly and shoot.

4. Upload the raw files.

You may want to do the same in room temperature and with a cold camera. I did not find the difference with the 40D relevant.
[a href=\"index.php?act=findpost&pid=181166\"][{POST_SNAPBACK}][/a]

Sorry, I don't test cameras that way, it's just experience with the 1ds mark 2 which I always found to be a little soft, especially with micro contrast, I've never used nikon digital, it seemed the dpreview was saying there was a little more noise on the nikon but a sharper file.

[Ray]

You are probably now aware that dpreview has just posted a full review of the D300. The results are pretty much in line with my own expectations, having already seen a few comparison images here and there on the interent.

Without getting in miniscule technical detail, the broad differences between the D300 and 40D seem to be that the D300 applies a default chroma noise reduction, whereas in the 40D it has to be enabled. The D300 also has several levels of noise reduction whereas the 40D has just one. The D300 generally has more options.

However, when it comes to comparing absolute image quality and noise characteristics at high ISO, the picture is not clear.

The D300 has an option for all noise reduction to be 'off'. Is this state comparable to the 40D's default position of NR disabled?

The D300 has a default NR called 'normal' which appears to apply an effective amount of chroma noise reduction, just as the 40D does when its one and only NR option is enabled. The noise reduction on the 40D, according to reports such as Bob Atkins' review, does not appear to degrade image resolution to any noticeable degree. This is a good characteristic of chroma noise reduction, as opposed to luminance noise reduction which does affect resolution.

I'll read dpreview's review again, but as I recall, I didn't see any comparisons between the 40D with NR disabled (which is the default position) and the D300 with NR disabled (which is not the default position).

Nor did I see any comparisons between these two cameras with default D300 NR (normal) and 'enabled' 40D NR, which is presumably similar.

In general, I find the image quality differences between the new breed of Nikons and Canons somewhat blurred by these different approaches to in-camera noise reduction.

[Panopeeper]

Ray,

I do realize, that not everybody is recording raw. I post this only because I read rumours elsewhere, that the NR of the D300 affects the raw data.

Neither the D300 nor the 40D apply noise reduction on the raw data. I.e. NR is irrelevant for raw shooters.

The only camera I know of, which applies NR on the raw data is the Sony A700.

[Ray]

Ray,

I do realize, that not everybody is recording raw. I post this only because I read rumours elsewhere, that the NR of the D300 affects the raw data.

Neither the D300 nor the 40D apply noise reduction on the raw data. I.e. NR is irrelevant for raw shooters.

The only camera I know of, which applies NR on the raw data is the Sony A700.
[a href=\"index.php?act=findpost&pid=181281\"][{POST_SNAPBACK}][/a]

Gabor,
Again, this issue is not clear. User adjustable noise reduction settings may not apply to the RAW image, but it's been widely reported that Nikon does apply chroma noise reduction to the RAW image, above a certain ISO level. And Nikon owners, of course, claim that Canon also apply noise reduction to their RAW image.

Are these just rumours? When I did a comparison a while back between the 5D and the D3, I got the impression that the 5D images after conversion were able to be improved to a greater extent, by applying 'chroma only' noise reduction in Noise Ninja, than the D3 conversions, the implication being that the D3 has already applied chroma noise reduction to the RAW image, whereas the 5D hasn't.

Unfortunately, dpreview have only compared the D300 and 40D at base ISO in RAW mode.

[sojournerphoto]

Gabor,
Again, this issue is not clear. User adjustable noise reduction settings may not apply to the RAW image, but it's been widely reported that Nikon does apply chroma noise reduction to the RAW image, above a certain ISO level. And Nikon owners, of course, claim that Canon also apply noise reduction to their RAW image.

Are these just rumours? When I did a comparison a while back between the 5D and the D3, I got the impression that the 5D images after conversion were able to be improved to a greater extent, by applying 'chroma only' noise reduction in Noise Ninja, than the D3 conversions, the implication being that the D3 has already applied chroma noise reduction to the RAW image, whereas the 5D hasn't.

Unfortunately, dpreview have only compared the D300 and 40D at base ISO in RAW mode.
[a href=\"index.php?act=findpost&pid=181455\"][{POST_SNAPBACK}][/a]

Slightly OT, but I think Chuck Westfall stated somewhere that Canon software hasn't applied any NR to RAW data since the introduction of CR2 RAW format.

Mike

[Panopeeper]

this issue is not clear. User adjustable noise reduction settings may not apply to the RAW image, but it's been widely reported that Nikon does apply chroma noise reduction to the RAW image, above a certain ISO level

First of all, there is no sense to talk about "choma" or "luminance" noise reduction at that stage. There are no two kinds of noise.

Second, Nikon cameras are applying the black levels correction (dark current noise), always, and this is not subject of discussion, as it is trivial. Canon cameras leave this to the raw processor, which is an inherent advantage of Canons (it is not Canons fault, that some raw processors, particularly ACR, suck in this point).

So, the question is only if Nikon applies any noise reduction *apart the above*.

Btw, Ray, do you happen to have some 5D raw images with problematic noise? Please upload some, I am just working on this area with Rawnalyze, and I need samples from other cameras as well (and opinions on the result).
[/quote]

[John Sheehy]

And if Adobe put some advanced NR in there, you may no longer see a difference. Who knows? Without extensive analysis of the raw data, it's hard to tell, but we do know Nikon put NR on the raw files before recorded,
[a href=\"index.php?act=findpost&pid=181152\"][{POST_SNAPBACK}][/a]

Where did you see that?

Are you talking about the fact that Nikon clips both the D300 and D3 above black in the RAW data?

I haven't noticed any signs of filtering in the RAW data.

[John Sheehy]

There is nothing new here. Canon leaves the "black current" in the raw data and it's the task of the raw processors to remove it. Nikon records the "amended" data.
[a href=\"index.php?act=findpost&pid=181166\"][{POST_SNAPBACK}][/a]

That's not black current.  I think you're talking about horizontal- and vertical-line-based
blackpoint offsets.

As for ACR ignoring them, I doubt it.  They were the first ones to actually compensate for it.  Did they just stop compensating?

[Panopeeper]

As for ACR ignoring them, I doubt it

I know for fact, that it does not, and I have never seen anyone claiming that.

[bjanes]

Slightly OT, but I think Chuck Westfall stated somewhere that Canon software hasn't applied any NR to RAW data since the introduction of CR2 RAW format.

Mike
[{POST_SNAPBACK}][/a]

That would depend on how you define noise reduciton. Canon does use on chip noise reduction and this is described in a [a href=\"http://www.robgalbraith.com/public_files/Canon_Full-Frame_CMOS_White_Paper.pdf]Canon Full Frame Whitepaper[/url] on page 17. This NR occurs on chip and before the raw data is written to memory. Unlike the filtering used to remove random noise at a later stage and that also results in a loss of image detail, these on chip methods reportedly do not cause loss of detail.

I would presume Nikon is using similar technology.

Bill

[BernardLanguillier]

However, when it comes to comparing absolute image quality and noise characteristics at high ISO, the picture is not clear.
[a href=\"index.php?act=findpost&pid=181275\"][{POST_SNAPBACK}][/a]

The conclusion of Phil on DPreview was pretty clear though, he prefers the way Nikon handles noise since he feels that it leaves more control to the user.

Cheers,
Bernard

[Ray]

Btw, Ray, do you happen to have some 5D raw images with problematic noise? Please upload some, I am just working on this area with Rawnalyze, and I need samples from other cameras as well (and opinions on the result).
[{POST_SNAPBACK}][/a]

Sending a few will be difficult. I'm back on a dial-up connection at present. I refuse to lock myself into a 24 month contract and pay the exhorbitant broadband fees in Australia when I intend spending a lot of time away from base anyway.

But here's one very noisy 5D RAW image that took me over an hour to upload. This is not ISO 6,400. It's not ISO 12,800 or even ISO 25,600. It's at least ISO 64,000.

It's one of a series of 9 shots I took for the purpose of experimenting with noise removal through stacking. It's really bad. Lots of banding   .

[a href=\"http://www.yousendit.com/download/www/a0YwbGtERnc1bmpIRGc9PQ]http://www.yousendit.com/download/www/a0YwbGtERnc1bmpIRGc9PQ[/url]

[attachment=5610:attachment]

[Panopeeper]

Thanks Ray, I downloaded it and am experimenting with it. It needs +4 EV, which makes it a pseudo-equivalent of ISO 51200.

ACR shows everything orangy with "as shot". Where would you pick WB? Or, if you adjust it as you remember to have seen it, what is the temperature and tint in ACR?

[John Sheehy]

The conclusion of Phil on DPreview was pretty clear though, he prefers the way Nikon handles noise since he feels that it leaves more control to the user.

[a href=\"index.php?act=findpost&pid=181880\"][{POST_SNAPBACK}][/a]

How does the user regain fine chroma detail filtered or desaturated by the converter?