Go Back   Steve's Digicams Forums > Digital SLR and Interchangeable Lens Cameras > Canon EOS dSLR

Reply
 
Thread Tools Search this Thread
Old Jun 9, 2007, 12:52 PM   #1
Junior Member
 
ripmcmanus's Avatar
 
Join Date: Jun 2007
Posts: 8
Default

I have collected several Canon digicams over the years, andstill use them all- G2, A620, Powershot Pro 1, A630, and Digital Rebel XT. I see lots of posts here and elsewhere about focusing issues with the XT, and I sharetheir concerns, but my theory is that the problem is not with the auto-focusing mechanism. Instead, I suspect the problem is overly aggressive noise reduction, and no doubt worseon the XTi than the XT.

Pictures coming from my XT (EFS 17-85 f4-5.6 IS) are almost always in focus (the plane at which I was pointing is the most focused part of the picture), but never have theclarity I see in pictures taken under identical conditions with the G2, Powershot Pro 1, or even the A620. Poorly lit background areas are definitely noiser on the older cameras, but that problem doesn't impact my subjective impression of the shot nearly as much as an object in the focal plane that lacks definition.

I was an audiophile "back in the day" and took issue with most noise reduction technologies at the time because they compromised upper octave frequency and phase response (clarity) in order to eliminate hiss. It seems to me that later generation Canons have an analogous problem with images. Yes, the image processing reduces noise dramatically, but there's a lot of valid image information lost in the process.

Thoughts? Ideas on how to avoid this issue, assuming it's true? Or am I all wet? Thanks in advance for any discussion.

Rip
ripmcmanus is offline   Reply With Quote
Sponsored Links
Old Jun 9, 2007, 1:21 PM   #2
Senior Member
 
aladyforty's Avatar
 
Join Date: Jan 2004
Posts: 6,964
Default

I was under the impression that noise removal programmes had improved a lot over the years and I see little problem with using them for some of my higher ISO shots. I do not use them excessively but when I do its hardly noticeable.
aladyforty is offline   Reply With Quote
Old Jun 9, 2007, 7:37 PM   #3
Super Moderator
 
peripatetic's Avatar
 
Join Date: Nov 2004
Posts: 3,599
Default

The answer is of course to shoot RAW; that will allow you to test.
peripatetic is offline   Reply With Quote
Old Jun 10, 2007, 12:11 AM   #4
Junior Member
 
ripmcmanus's Avatar
 
Join Date: Jun 2007
Posts: 8
Default

Absolutely. I'll give RAW a shot. To the response above, I'd just say it's not that noise reduction isn't working. Clearly it works. But it works too well. The problem is that real image detail can look just like noise and be eliminated right along with it. As a logical extreme, consider taking a perfectly-focusedpicture of a high-res image of camera noise. In theory, with "perfect" noise reduction, the noise-reduced image would be almost 100% inaccurate.

For the heck of it (and I will compare RAW as soon as I can), I've attached a small side-by-side comparison of croppings from pictures of one of my cats, a Rag Doll with fine, long hair. This seems like a reasonable challenge for noise reduction algorithms to me, because there is only a small amount of contrast in rather large areas of the cat's hair. The Pro1 shot was taken at f3.2, 1/60th sec, ISO 100, built-in flash. The XT shot was f4.0, 1/60th sec, ISO 400, Speedlite 420EX external flash. Both were shot at highest resolution and lowest compression jpeg. I believe these conditions are roughly comparable or favor the XT slightly. The Pro1 maxes out at ISO 400 and reviews noted noise at settings as low as ISO 100. The difference is easily visible (to me) in these two samples, but it's actually much more pronounced in the context of each fullpicture (both of which are muchtoo large to upload).

As Canon attempts to squeeze ever more pixels out ofvery small sensors, noise reduction algorithms must necessarily become more aggressive. I'm going to check RAW to see what I get, but I'm not thrilled about the prospects of cutting the number of shots I can take by 80% to accommodate the difference in file size, or the amount of additional time it will take to process each image to something pleasing to the eye. For now, the XT sits mostly unused on my desk except in the most extreme shooting conditions.
Attached Images
 
ripmcmanus is offline   Reply With Quote
Old Jun 10, 2007, 1:41 AM   #5
Administrator
 
Join Date: Jun 2003
Location: Savannah, GA (USA)
Posts: 22,378
Default

ripmcmanus wrote:
Quote:
The Pro1 shot was taken at f3.2, 1/60th sec, ISO 100, built-in flash. The XT shot was f4.0, 1/60th sec, ISO 400, Speedlite 420EX external flash. Both were shot at highest resolution and lowest compression jpeg. I believe these conditions are roughly comparable or favor the XT slightly.
You're testing it with the aperture wide open, and most lenses designed for SLR or DSLR cameras are going to be a bit softer that way. Try f/5.6 or f/8 on the wide end of that lens and it should sharpen up some.

I doubt what you're seeing has anything to do with noise reduction algorithms. It's more likely from other causes.

Are you sure about the in focus part?

You're going to have a much shallower depth of field for any given focus distance, angle of view and aperture with a DSLR (because the actual focal length of the lens on a camera like your Pro 1 is much shorter for any given angle of view due to it's much smaller sensor). So, you're going to need to be more careful about where you lock focus (or stop down the aperture more for greater depth of field if you're shooting a closer subject).

Also, a DSLR is not going to process the image as aggresively as a model like the Pro 1 (you may need to increase sharpness, saturation, contrast and more if you want the same "look" as you're getting with a non-DSLR model). For best results, use an editor later so that you can apply as little or as much as desired for purpose of the image.

Quote:
As Canon attempts to squeeze ever more pixels out ofvery small sensors, noise reduction algorithms must necessarily become more aggressive.
Your Rebel XT has a MUCH larger sensor compared to your Pro 1.

JimC is offline   Reply With Quote
Old Jun 10, 2007, 1:41 AM   #6
Junior Member
 
ripmcmanus's Avatar
 
Join Date: Jun 2007
Posts: 8
Default

I just did a quick jpeg vs RAW comparison, although not on the cat (didn't want to wake her). The RAW was processed by Zoombrowser EX, which provided no option to determine the amount of noise reduction to apply. The controls provided duplicate those in the camera, but I actually reduced the contrast from its normal Parameter 1 default before converting.Even so, the jpeg created from the RAW conversion process is definitely somewhat clearer than the jpeg produced by the camera at the same time.

In the pics below, I can pretty clearly see horizontal lines in the pink area in the jpeg converted from RAW that are almost invisible in the jpeg created by the camera (presumably eliminated by noise reduction). There is also more pronounced "noise" in the green leaf- small, almost patterned differences in color that are actually present in the object that was photographed.

I'm still annoyed by the need to use 14mb for each shot instead of 3.7, and at the amount of additional time it takes to process each image, but for important shots, RAW certainly seems like the way to go. Meanwhile, I also think this somewhat unscientific sample lends credence to the argument that noise reduction is not always our friend!
Attached Images
 
ripmcmanus is offline   Reply With Quote
Old Jun 10, 2007, 2:02 AM   #7
Administrator
 
Join Date: Jun 2003
Location: Savannah, GA (USA)
Posts: 22,378
Default

Quote:
In the pics below, I can pretty clearly see horizontal lines in the pink area in the jpeg converted from RAW that are almost invisible in the jpeg created by the camera (presumably eliminated by noise reduction). There is also more pronounced "noise" in the green leaf- small, almost patterned differences in color that are actually present in the object that was photographed.
That's more likely sharpening differences in the algorithms used for raw conversion. Try sharpening the camera produced jpeg. ;-)

JimC is offline   Reply With Quote
Old Jun 10, 2007, 2:34 AM   #8
Junior Member
 
ripmcmanus's Avatar
 
Join Date: Jun 2007
Posts: 8
Default

Excellentpoint about sensor size! But it's still true that the Pro 1 does not employ nearly as much in-camera noise reduction as the Rebel XT, and that's my point. Although the Pro 1 exhibits much more noise when I zoom in on areas of the picture Ioften don't care about, it hasn't "edge-preserving smoothed" away subtle details in areas of the picture I do care about (and that I'm convinced many Rebel owners have mistaken for problems with auto-focus).

Your point about doing more of the image processing inPhotoshop, et al, is right on target. I just wish it didn't take shooting in RAW, waiting 3 seconds for each shot to be written to the CF card,and spending several minutes tweaking every picture to get my money's worth out of the Rebel XT. And since I seldom use any of my cameras in a studio setting, I'm resigned to the fact that if I want the highest quality unposed images with the least amount of work the highest percentage of the time, the Pro 1 or A620 is my best bet by a country mile.

I certainly agree with your recommendation to use higher f-stops to increase depth of field when appropriate, butthere's not much wiggle roomwhen I'm shooting indoors. Even with a flash that cost more than some cameras, the best the Rebel XT could dial up was ISO 400 at 1/60th of a second at f4. A higher f-stop would necessitate an ISO or exposure setting that I suspect would have resulted in an even less satisfying image. And don't forget that the f3.2 setting dialed in by the Pro 1 isn't far from its most wide-open aperture setting, either!
ripmcmanus is offline   Reply With Quote
Old Jun 10, 2007, 2:47 AM   #9
Super Moderator
 
peripatetic's Avatar
 
Join Date: Nov 2004
Posts: 3,599
Default

Hmm, don't forget that with flash you can go down to ISO 100 if you like (obviously there are lots of factors involved with flash, but if your flash can light the scene then there is no reason to use high ISO).

14MB is pretty high for the XT RAW file, it should normally run around 8 or so.

JimC correctly points out that there may be significant differences in depth-of-field between the XT and the Pro1, with the same angle-of-view (framing) the Pro1 will generally have much greater DOF.

Another possibility is that you might be experiencing lens issues. The lens on the Pro1 is an L lens, your XT lens may be considerably less sharp. What lens are you using on the XT, and where in the frame were these crops taken from?

It would help if you could post a reduced version of the whole shot along with the crops.

Another thing that you might like to try out is to see whether DXO magically solves your problems for you.

Download a trial version and play with it a bit, it knocks the socks off any in-camera algorithm and allows you to easily batch process all your shots.

http://www.dxo.com/intl/photo/free_demo
peripatetic is offline   Reply With Quote
Old Jun 10, 2007, 3:11 AM   #10
Junior Member
 
ripmcmanus's Avatar
 
Join Date: Jun 2007
Posts: 8
Default

I definitely understand what you're suggesting, and I'm keeping an open mind!

BUT...the camera was set at +1 sharpening, sothe JPEG produced by the camera (JPEG with RAW) reflects that setting. The JPEG produced by the RAW conversion in Zoombrowser EX (JPEG from RAW) was generated with the same setting on a RAW image that wasn't sharpened in the camera, so if Canon uses the same algorithm in the XT that they use in Zoombrowser, the results should be identical. They aren't.

To see if the differences were due to the sharpening algorithm or the RAW format, I sharpened the same area of the JPEG produced by the camera (see below). To my eyes, it madeno difference, and I didn't expect it to since the camera had already applied excess sharpening. More to the point, it still lacks the clarity of the JPEG converted from RAW in Zoombrowser EX.

Since there seems to be some doubt that aggressive noise reduction can reduce clarity in an image, I also aggressively reduced the noise on the JPEG converted from RAW (also see below). Let's laythis one to rest - too much noise reduction CAN compromise an image's clarity!
Attached Images
 
ripmcmanus 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 1:50 AM.