Canon 7D vs Mark IV Noise levels when cropping is required

[Alan Stankevitz]

I have been using the Canon 7D for a few months now doing mainly bird photography with it. At first, I was a bit taken back by the higher-than-expected noise levels but eventually realized that noise levels were similar to the 5D Mark II when photographing birds from a set distance. In other words, the 5D Mark II images had to be more heavily cropped than the 7D due to sensor dimensions (1.6x vs full frame) when photographing birds from a set distance and the 7D outresolved the 5DM2 with about the same level of noise.

Since most of us, most of the time are unable to walk up to a bird I am wondering if the extra cropping required by the Mark IV will negate the benefits of lower noise levels.

Your thoughts?

Thanks,

Alan

[Doug Brown]

If it's anything like the Mark III, it won't negate the benefits of lower noise levels. Mark III RAW files are stunning and from what I've seen of Mark IV RAW files, they are no less stunning.

[Roger Clark]

I have been using the Canon 7D for a few months now doing mainly bird photography with it. At first, I was a bit taken back by the higher-than-expected noise levels but eventually realized that noise levels were similar to the 5D Mark II when photographing birds from a set distance. In other words, the 5D Mark II images had to be more heavily cropped than the 7D due to sensor dimensions (1.6x vs full frame) when photographing birds from a set distance and the 7D outresolved the 5DM2 with about the same level of noise.

Alan,
I am confused by your statement. The 5D2 pixels are larger than 7D pixels thus have higher signal-to-noise (S/N) ratios. So images with the 5D2 have lower resolution on the subject but the pixels have higher S/N. If you
compare enlarged images where the subject is the same size on the print/display, then the noise per square mm would be similar for brighter parts of low ISO images, but the larger pixel camera images will show lower noise as the ISO increases and one examines the darker portions of the images. Crop factor has nothing to do with it. It is all a matter of pixel size and sensor size.

A couple of references:
http://www.clarkvision.com/articles/cropfactor

Does pixel size matter:
http://www.clarkvision.com/articles/...el.size.matter

Roger

[Alan Stankevitz]

Here's my point...

The 5D Mark II is touted as a low-noise camera...which it is when you can fill the frame with the subject your are photographing...portrait photography, etc.

In my test, I have taken images from both the 5D Mark II and 7D from a set distance using the same lens at ISO 1600 for a test. The subject (text on a box) takes up about 1/6 of the frame on the 7D while the 5DM2 takes up only about 1/10 of the frame. (Similar to what you might expect when photographing a perched bird from a distance.)

Once both images were loaded into Photoshop, I cropped each image to 300 ppi so that the box with the text fills the frame by about 1/3rd in both cases.

The 5DM2 image shows slightly less noise with lower resolution when compared to the 7D...they are very close however.

(If you care to look at the examples, you can view the images here. They were both converted from RAW w/o any noise reduction:

7D:
http://www.iwishicouldfly.com/iwishi...ped_300ppi.jpg

5DM2:
http://www.iwishicouldfly.com/iwishi...ped_300ppi.jpg

My point is that the 1.3x sensors and full frame sensors require greater cropping of the subject in order to create a properly cropped print. In doing so, noise is amplified and this somewhat negates the "lower noise" touted in 1.3x and full-frame sensors. That is why I am wondering if the Mark IV will truly perform any better (noise-wise) than the 7D when photographing from a set distance to a subject. My guess is that it will, but only by a small margin.

Again, this only applies to situations in which your are focal-length limited...in other words you must shoot from a set distance from your subject.

Alan

[arash_hazeghi]

Alan,
I think a better title for this thread is overall IQ when you are focal length limited.
When cropped to 1.6X FOV MKII will only have 8 Mpixels compared to 7D's 18 Mpixels. Of course up-sampling the 5DMKII files to match that of the 7D crop will exaggerate appearance of noise due to software extrapolation and the coarse grainy look that results, when evaluating noise you have to look at 1:1 sensor output for fair comparison, in this case you need to use a longer lens on a 5DMKII.

Back to your question, which is an interesting discussion. IMO the answer is that it really depends and it isn't as straightforward as shooting a text target on tripod in good light like your examples. It depends on subject and its size, light conditions, optics and your post processing skills. I have a Canon 7D which is my primary BIF camera, a 5DMKII which I use for landscape and large prints and a Nikon D700 which I have also tried birding with. I can confidently claim I have had some situations where the low resolution Nikon delivered results which would not have been possible with 7D despite its reach advantage, the feather details would have completely "sunk" in noise beyond recovery. The other issue is optics, smaller pixels demand better optics and slight miss-focus shake or other imperfections will degrade IQ. These are less of an issue for large pixel cameras. So at the end of the day there is no black and white answer, in general I have experienced that if you shoot a lot of smaller birds in situations that you do not need to be above ISO 800 (like perched songbirds or shorebirds in good light) small pixel cameras such as 7D do offer a big advantage when you are focal length limited. In low light situations (ISO 1600 and higher for example) with birds that have low contrast feathers there is not much advantage from the smaller pixels and you would be better off using a large pixel camera and cropping tighter. Of course if the bird ends up being so small that you can't pull anything out, you need a large pixel camera AND a longer lens! :eek: not a smaller pixel camera. In these cases I have no doubt that 1DMKIV or even current MKIII will provide superior results. On paper a MKIV 1.3X 16mpixel sensor can be a very good compromise between noise and resolution providing descent results perhaps up to ISO 3200. In some situations 7D will deliver better overall IQ due to its reach but in some other MKIV's larger pixels will win.

[Doug Brown]

Alan,
I think a better title for this thread is overall IQ when you are focal length limited.

I don't think he's talking about focal length. It has more to do with a fixed distance to subject.

[Roger Clark]

I agree with Arash, and good description on the resampling issues with noise. One thing that strikes me on these test images is how close the resolution is between the 5DII and 7D. The 7D images also show less contrast, which is also predicted by the modulation transfer function. The reduction in contrast is an indication of loss of detail.

If the test were redone at higher ISO, the 7D would be degraded more. An a test of text does not show the fine detail that a feather test would show.

Roger

[Joel Eade]

Roger,
It seems that every camera has a prominent weakness. Either resolution, low light performance, frame rates or some other feature that comes up just short of what would be optimum. Due to the limits physics is it impossile to make a camera with all the deisred features? Based on your scientific analysis and experience ... using on present day technology what properties would the optimum sensor/camera have? Or are we just asking too much?

[Alan Stankevitz]

Same test but with feathers:

5DM2:
http://www.iwishicouldfly.com/iwishi...pi_feather.jpg

7D:
http://www.iwishicouldfly.com/iwishi...pi_feather.jpg

In the previous test, I did let ACR auto adjust exposure. This time nothing has been altered except color noise reduction setting of 25 for both images.

Both images were taken with the Canon 600mm f/4 IS lens at a set distance of 30'.

I suspect the 1DM4 will have slightly better resolution than the 5DM2, but not as good as the 7D. I'm sure the noise level will be lower however and how much of a role it plays towards overall image quality will be the key.

Canon has done such an excellent job with the 7D's autofocus and fps, the advantage of the 1DM4 will boil down to its noise level capability. So far from what I have seen on the net, it is quite good, but from a bird photographer's viewpoint where in many instances we are cropping our shots, how much better will it be? The more you crop an image, the more noticeable the noise level becomes. A 1.3x sensor will require a heavier crop than a 1.6x sensor.

For those of you who have the 1DM4 on order, it would be great to do some comparison cropping between the 1DM4 and the 7D and post them here!

Alan

[arash_hazeghi]

I don't think he's talking about focal length. It has more to do with a fixed distance to subject.

True Doug, I think that is what "focal length limited" usually means: fixed distance and far subject when you have only one lens.

[Alan Stankevitz]

Arash...that's correct. A focal-length-limited situation is when you cannot physically move closer to the subject using a particular lens. In my case, I quite often shoot using a 600mm with a 1.4x TC (840mm) and I still have to do some cropping.

In situations (such as in Florida) where birds are tame, the 5D Mark II is a better option for me filling the frame with nothing but feathers!

[Emil Martinec]

Here's my point...

The 5D Mark II is touted as a low-noise camera...which it is when you can fill the frame with the subject your are photographing...portrait photography, etc.

In my test, I have taken images from both the 5D Mark II and 7D from a set distance using the same lens at ISO 1600 for a test. The subject (text on a box) takes up about 1/6 of the frame on the 7D while the 5DM2 takes up only about 1/10 of the frame. (Similar to what you might expect when photographing a perched bird from a distance.)

Once both images were loaded into Photoshop, I cropped each image to 300 ppi so that the box with the text fills the frame by about 1/3rd in both cases.

The 5DM2 image shows slightly less noise with lower resolution when compared to the 7D...they are very close however.

(If you care to look at the examples, you can view the images here. They were both converted from RAW w/o any noise reduction:

7D:
http://www.iwishicouldfly.com/iwishi...ped_300ppi.jpg

5DM2:
http://www.iwishicouldfly.com/iwishi...ped_300ppi.jpg

My point is that the 1.3x sensors and full frame sensors require greater cropping of the subject in order to create a properly cropped print. In doing so, noise is amplified and this somewhat negates the "lower noise" touted in 1.3x and full-frame sensors. That is why I am wondering if the Mark IV will truly perform any better (noise-wise) than the 7D when photographing from a set distance to a subject. My guess is that it will, but only by a small margin.

Again, this only applies to situations in which your are focal-length limited...in other words you must shoot from a set distance from your subject.

Alan

Before drawing conclusions from your two sample images, I'd like to ask a few pertinent questions:

1) What converter was used (and if ACR, what dot release -- it matters)?
2) In real life, is the green patch on the right side of the lamp bulb box corrugated as in the 7D image, or smooth more like the 5D image?
3) During conversion, what manipulations were made -- any NR, exposure compensation, etc?

[Alan Stankevitz]

Hi Emil,

1) I used vers. ACR 5.6
2) The green patch on the side of the box is made up of small dots. I used this box because it was the smallest text I could find at hand.
3) A level of 25 was used for color noise reduction, luminance noise reduction was set to 0, sharpening = 0, exposure compensation = +.75

Alan

[Alan Stankevitz]

Browsing Flickr.com I found comparison photographs between the 7D and Mark IV at various ISO's taken from the same distance (w/o noise reduction applied).

Cropping the photographs to the exact same frame reveals that the 7D has more detail but with more noise. By applying NR to the 7D cropped image to equal similar noise levels of the Mark IV, I get about the same level of detail.

This is by no means conclusive and frankly I'm getting tired of seeing just night/dark images taken with the Mark IV at ridiculously high ISO's!

Here are the links to the comparison images:

-- 7D:

http://www.flickr.com/photos/hirokik...439495/detail/

-- 1D Mark IV:

http://www.flickr.com/photos/hirokik...437379/detail/

Special thanks to Hiroki in Japan for these images.

Cheers,

Alan

[Aaron Lee]

In my case, I quite often shoot using a 600mm with a 1.4x TC (840mm) and I still have to do some cropping.

Alan,greetings-I assume you are shooting truly wild birds in the country or a woods?

[Alan Stankevitz]

Hi Aaron,

Yes, that is correct. Mostly in the Midwest U.S. It is rare that I run across a bird that is tame enough to get close. Using a blind does help, but not always practical.

Alan

[Emil Martinec]

Hi Emil,

1) I used vers. ACR 5.6
2) The green patch on the side of the box is made up of small dots. I used this box because it was the smallest text I could find at hand.
3) A level of 25 was used for color noise reduction, luminance noise reduction was set to 0, sharpening = 0, exposure compensation = +.75

Alan

OK, thanks. The reason I ask is that the 7D has a substantial problem with its RAW data, a lot of vertical banding even at low ISO, that requires special attention in the converter programming. ACR 5.5 was not specially programmed and so had all sorts of artifacts; ACR 5.6 included changes to the demosaic for the 7D which helped a lot, however one can still see a residue in the green patch on the right part of the box, where small dots become fine vertical bands. There is also a more "vertical" character to the noise in OOF areas. Because of these shortcomings of the 7D, I wouldn't necessarily make sweeping conclusions about small pixels vs large. Canon cameras of late have been rather hampered by pattern noise issues.

I disagree with Arash and Roger that the proper method of evaluation is always a 100% view, as this risks losing the forest for the trees with high resolution sensors. I have yet to see a demonstration of details that can be captured by a lower resolution that are too "sunk" in noise to be captured by a higher resolution sensor. I have seen plenty of examples of the opposite -- details that a lower resolution sensor cannot resolve that a higher resolution sensor can. That is especially true when details near the pixels scale (the Nyquist frequency), where demosaic artifacts become more problematic with coarser pixels.

It is also misleading to characterize high resolution sensors as being somehow worse performers due to lens optics, camera shake and the like. Rather they perform as well or better than coarse resolution sensors, they simply resolve better and therefore all the imperfections in lenses and technique are simply more accurately revealed than smeared away by poor resolution. And you always have the option of smearing them away if you want.

I think that high resolution sensors would benefit from a renovation of software algorithms, which were designed in an era of low resolution sensors. Many demosaic algorithms are too easily confused by pixel scale noise; downsampling algorithms too often alias high-frequency noise into the output image (PS bicubic is an unfortunate case in point) and thus show more noise than they might in the downsampled image. Fortunately noise filters are showing some improvement.

[Alan Stankevitz]

Emil,

I have been critiquing different RAW converters this past week. ACR, DPP, Bibble 5 Pro and a few others. Since I currently own ACR, I really don't want to invest in another converter, but if it does a better job, I will. Do you have any preferences?

The latest release of ACR (5.6) OK and DPP works fine as long as you don't use there noise reduction. I have found DPP's noise reduction to leave blurred areas blotchy.

I liked Bibble Pro 5 quite a bit with there incorporation of Noise Ninja.

On another note, I did some crop sampling (similar to my test) with Hiroki's images and found the Mark IV images to be better noise-wise than the 7D, but with lower resolution. I found that when I reduced the noise in the 7D image to equal that of the Mark IV, the resolution between both cameras was pretty much a wash.

This of course, is just from jpg samples. I'd like to do the same with RAW images from both to get an accurate account of noise levels vs resolution.

Cheers,

Alan

[Roger Clark]

Roger,
It seems that every camera has a prominent weakness. Either resolution, low light performance, frame rates or some other feature that comes up just short of what would be optimum. Due to the limits physics is it impossile to make a camera with all the deisired features? Based on your scientific analysis and experience ... using on present day technology what properties would the optimum sensor/camera have? Or are we just asking too much?

Joel,
I think it is possible to make the ultimate camera. The problem is if camera manufacturers did that, their sales would drop, once everyone (who is going to buy one) buys one. Bill gates taught the world a "better" business model (he stated this in an interview with Barbara Walters): sell a defective product so you can sell an upgrade (not an exact quote), but the model seems to be practiced by many companies these days. Fortunately, each iteration gets a little closer. Perhaps some of these new features are experiments to see how they work. In particular, AF systems are simple in concept and difficult in practice and even more so as pixel pitch decreases. Mechanical tolerances become extremely small, especially for a moving part (the mirror) and then the AF must focus on what the photographer wants in possibly complex backgrounds or foregrounds; no small feat. Sensors are pretty mature, but refinements are still happening, allowing more megapixels processed at a faster rate. The good news, is that cameras are getting and will continue to get better, but increments will be smaller and smaller, trying to milk every dime from us. (Cynicism off)

Roger

[Roger Clark]

Here is a way to look at the problem, see Apparent Image Quality, AIQ, in Figure 9 at:
http://www.clarkvision.com/articles/...index.html#AIQ

When you crop, you effectively have a smaller sensor. So find the 5D Mark II point and move it down to the 1.6x-crop sensor line (the red dashed line). Then you would see the 5D2 plots below the 7D in terms of the AIQ model. Your test images were at f/4 on a great L lens, so the resolution is delivered by the lens. Stop the lens down to f/8 and the 7D results will be degraded and only slightly better than the 5DII. As light levels drop and your pump of the ISO high, the higher noise of the 7D will harm the image more. Although I am impressed by your test with ISO 1600 images, it is an easy target with a very small dynamic range. The 7D has a stop less dynamic range than the 5D2 (per pixel) and in more extreme lighting that will become an additional factor, but it will be small (the smaller pixels also helps narrow the difference)

Personally, I think the ideal camera is a full frame camera with about 5-micron pixel pitch, 10 frames/sec, around 35 megapixels. We are getting close.

Roger

[arash_hazeghi]

I did a quick test with 7D and MKII (don't have the right lens for D700) with Canon 70-200 f/2.8 IS (which is one of Canon's best lenses) on tripod and LV manual focusing + remote control shooting. This is NOT a scientific test or noise measurement etc. just an quick "overall" IQ comparison, RAW files converted with DPP 3.7.2 which does the best for any EOS camera. I used sharpening =3, chroma NR=8 and Luminance NR=0 settings which is my standard workflow. lens set to 70mm and f/8 ISO 1600, this scene is somewhat low light (f/8 1/10sec) compared to Alan's but still uniformly lit so not a DR challenge.



As you can see features are larger in 7D image but they contain no extra detail that is visible to my eyes, maybe someone can point out if they actually see more detail in 7D image? Grain has interfered with detail a bit, not bad for 7D but no amount of down-sampling, noise reduction etc. will make it look better than the MKII output in this case.
For me a 5D MKII sensor in a 7D body would have been an ideal camera, with a 500 and 1.4X TC I get enough reach for my subjects that are raptors and shorebirds and I could also take wide angle photos and enjoy low noise + good DR in one package. IMO Canon could have had a large market for such camera, just like Nikon has been selling the D700 like pancakes.

[Alan Stankevitz]

Arash,

Maybe you missed my point of my original post? The test involved cropping an image identically between two different cameras. I did this because when I photograph a perched bird for example, I crop the original photo to highlight the bird. When I crop, I typically crop to 300 ppi since that is commonly what publications want. For example a full-frame sensor would require heavier cropping than a 1.6x sensor in order to crop identically.

The example you show is not cropped identically, which while interesting does not follow my reason for posting. Also, I would suggest turning off noise reduction and no sharpening when comparing.

Cheers,

Alan

[Alan Stankevitz]

Roger,

You may have a point regarding dynamic range. That does make a difference from a quality standpoint.

Thanks,

Alan

[arash_hazeghi]

Arash,

Maybe you missed my point of my original post? The test involved cropping an image identically between two different cameras . I did this because when I photograph a perched bird for example, I crop the original photo to highlight the bird. When I crop, I typically crop to 300 ppi since that is commonly what publications want. For example a full-frame sensor would require heavier cropping than a 1.6x sensor in order to crop identically.

The example you show is not cropped identically, which while interesting does not follow my reason for posting. Also, I would suggest turning off noise reduction and no sharpening when comparing.


Cheers,

Alan

Alan,

I understand what your point is, this is exactly the demonstration of focal length limited case where I have shot a scene with the same lens from a fixed distance, if you haven't noticed it is exactly the same crop, therefore objects are precisely 1.48X times larger in 7D photo compared to 5DMKII [(sqrt(18Mpixel/21Mpixel)*1.6=1.48]. For a given print size, you can up-sample the MKII image or down-sample the 7D image to see for yourself if 7D has any advantage in these conditions.

When photographing birds in flight we mostly down-sample images for web presentation, so you can down-sample both images to your desired size and see which one has more detail.

Also turning sharpening off or on has no effect on the relative performance as long as both images are processed identically, best way is to run the images through your standard work flow which is what matters at the end of the day. I am not sure how you processed the crops you posted but they seem to have a lot of digital artifacts, maybe you enlarged both of them?

Here I have up-sampled MKII image to match the FOV of 7D, as you can see it looks a bit coarse in detail but still it has better definition to my eyes, the 7D image is grainier and can give you the illusion of more detail or sharpness but if you look carefully the edges of text is better defined in MKII image. In this case MKII will produce a better print at any given size.



Remember that this is still good light and ISO 1600 in low light and ISO 3200 MKII advantages will be more pronounced. Any ways If you shoot birds in flight in various different conditions you will sure experience these cases for yourself in practice :)


Good luck

[Aaron Lee]

Arash,
1.Why did you use ISO 1600 off a tripod?
2.Was the high ISO noise reduction "on" on the 7D? [that will decrease detail]
3. In the Mark II image I can see the Appalachian trail,the 7D image has more detail-I can see Mark Sanford on one of the trails.:D

[Emil Martinec]

Arash,

IMO both conversions suffer from a lot of demosaic artifacting, which is one of the biggest shortcomings of DPP. These artifacts limit the ability to sharpen the image, and obscure any resolution advantage conferred by the higher density sensor. I would also venture to say that Canon has not dealt as well with the banding issues in the 7D raw data as Adobe has with ACR 5.6, and that is a substantial source of the greater artifacting seen in the 7D conversion.

Alan,

It's been a while since I last did a careful comparison of the available converters, and I don't own a 7D which as I've been mentioning has its own very particular issues. It seems that ACR 5.6 does about the best job of mitigating those issues that I've seen (and again that's other people's work; I skipped CS4 and am waiting for CS5 for my next upgrade). Hopefully the tweaks they made in going from ACR 5.5 to 5.6 which are specific to the 7D will make their way into LR3beta, which has a much improved demosaic engine relative to previous versions of ACR/LR.

[arash_hazeghi]

I don't believe DPP has any particular artifatcts except for the infamous 45 degree jaggies neither have I obsereved any banding issue in real world samples of my 7D. IMO DPP produces the best looking images of any EOS camera. Conversions with ACR look grainy and softer in comparissons that I have done.

[Emil Martinec]

I don't believe DPP has any particular artifatcts except for the infamous 45 degree jaggies neither have I obsereved any banding issue in real world samples of my 7D. IMO DPP produces the best looking images of any EOS camera. Conversions with ACR look grainy and softer in comparissons that I have done.

DPP generates quite a lot of zipper artifacts, as well as vertical/horizontal bars that are one pixel wide and several pixels long (usual 2-3). Your conversions show plenty of these little bars. You may be thinking of it as "grain", but it's interpolation error. The cause of the bars is a binary demosaic -- the missing colors (or rather the color differences G-R and G-B) are interpolated either entirely in the vertical direction or entirely in the horizontal direction, and this can generate little gridlike patterns that don't sharpen well. ACR does this too, it just tries to hide it in a softer conversion. Unfortunately I don't know of any current converter that does a proper adaptive interpolation (with the exception of DxO, which uses a wholly different approach to demosaic based on nonlocal means interpolation, and which has its own set of issues; also, since I don't use Nikon, I haven't looked at NX).

The banding is in the raw data if you look at it directly. On the worst copies of the 7D I've seen, it is about twice as large as photon shot noise, at all exposure levels, because it's a gain banding rather than an offset banding. Different converters suppress it to different degrees.

[arash_hazeghi]

DPP generates quite a lot of zipper artifacts, as well as vertical/horizontal bars that are one pixel wide and several pixels long (usual 2-3). Your conversions show plenty of these little bars. You may be thinking of it as "grain", but it's interpolation error. The cause of the bars is a binary demosaic -- the missing colors (or rather the color differences G-R and G-B) are interpolated either entirely in the vertical direction or entirely in the horizontal direction, and this can generate little gridlike patterns that don't sharpen well. ACR does this too, it just tries to hide it in a softer conversion. Unfortunately I don't know of any current converter that does a proper adaptive interpolation (with the exception of DxO, which uses a wholly different approach to demosaic based on nonlocal means interpolation, and which has its own set of issues; also, since I don't use Nikon, I haven't looked at NX).

The banding is in the raw data if you look at it directly. On the worst copies of the 7D I've seen, it is about twice as large as photon shot noise, at all exposure levels, because it's a gain banding rather than an offset banding. Different converters suppress it to different degrees.

I have no idea what you are talking about "one pixel bars" both files look normal to my eye they show some noise but nothing I can identify as a particular artifact. Anyways this is the converter that I use. Also no one except Canon knows exactly what algorithms they use I don't have this information and wouldn't read too much into it. Some FP noise is expected for high density sensor with multi channel high speed readout mainly due capacitive cross talk but it has no detremental effect on resolution in the examples I posted above and I have never seen it in real world photos, sensor fabrication process is very mature and yields are high so I wouldn't expect that much variation between cameras. Any ways I am happy with both 7D and MKII but I don't believe 7D has a real reach advantag over MKII in all shooting situations.

[Emil Martinec]

I have no idea what you are talking about "one pixel bars" both files look normal to my eye they show some noise but nothing I can identify as a particular artifact. Anyways this is the converter that I use. Also no one except Canon knows exactly what algorithms they use I don't have this information and wouldn't read too much into it.

I marked up a small crop and blew it up 4x to make it more obvious; I didn't mark them all, so that the image would still be visible ;) :



As for not knowing exactly what algorithm Canon uses, I'm not privy to that but I have enough experience in demosaic algorithm design to know from the output whether the interpolation uses a binary vertical/horizontal choice or instead an adaptively weighted choice. The latter provides a much more natural looking image, but as I mentioned it isn't currently used. Adobe products also suffer from a non-weighted interpolation, but they prefer a smoother initial output that can withstand more sharpening later in post, and thus iron out some of these issues.

[arash_hazeghi]

Again to me these don't look anything more than noise and sharpening artfifacts I see maybe a couple of brigth pixels in a single column that may have come from demosaic but I wouldn't care about them since they are irrelevant for any practical usage unless one looks at photos at 200%. It doesn't change the general arguments made above. If you prefer a more natural looking image you can always turn sharpening off and then sharpen in other Programs. Any ways thanks for posting.

[Emil Martinec]

Again to me these don't look anything more than noise and sharpening artfifacts I see maybe a couple of brigth pixels in a single column that may have come from demosaic but I wouldn't care about them since they are irrelevant for any practical usage unless one looks at photos at 200%. It doesn't change the general arguments made above. If you prefer a more natural looking image you can always turn sharpening off and then sharpen in other Programs. Any ways thanks for posting.

Well, the whole issue in this thread is the ability to extract detail. Demosaic artifacts provide false detail, which is not noise per se but interpolation error. It's not noise because it has a regular structure -- pixels are drawn from two sample distributions, with the odd pixels belonging to one sample and the even pixels to the other. The telltale signs are all over the 7D image especially. It does have effects similar to noise -- it intrudes on larger scales than the pixel level due to its impulsive nature, it limits one's ability to sharpen the image (either in DPP or elsewhere), etc.

You can't have it both ways -- argue that DPP is providing more detail (especially when it's false detail), and then say that it's irrelevant for practical usage because nobody looks at that level of detail.

[Alan Stankevitz]

Just as a note, today I photographed birds at one of my feeding stations from the same distance with both the 7D and 5DM2. Using f/8, the 7D outresolved the 5DM2 and had similar noise levels all the way up to ISO 1600. At ISO 3200, the 7D exhibited more color noise but still had greater detail.

Alan

[arash_hazeghi]

Well, the whole issue in this thread is the ability to extract detail. Demosaic artifacts provide false detail, which is not noise per se but interpolation error. It's not noise because it has a regular structure -- pixels are drawn from two sample distributions, with the odd pixels belonging to one sample and the even pixels to the other. The telltale signs are all over the 7D image especially. It does have effects similar to noise -- it intrudes on larger scales than the pixel level due to its impulsive nature, it limits one's ability to sharpen the image (either in DPP or elsewhere), etc.

You can't have it both ways -- argue that DPP is providing more detail (especially when it's false detail), and then say that it's irrelevant for practical usage because nobody looks at that level of detail.

As I stated I don't believe that you can extract more detail out of the 7D file under some conditions like the example I posted, no matter what converter is used and I also believe DPP is the best converter for current Canon cameras I don't really see any serious systematic issue, without detailed information about Canon's algorithms I would not comment or try to figure out where a few bright pixels in a row or column originated from and I don't think they will affect the extinct resolution of the camera. Of course opinions will vary on an Internet forum. The best way for the OP is to test these cameras in the field and see for himself which camera suits his needs better given his conditions and processing parameters. 5D MKII is not very suitable for birding due to its AF and speed so the detail argument is a bit irrelevant any ways Thanks for your input.
good luck

[Alan Stankevitz]

Arash and others,

At least the 7D that I own has got the best resolving power of any of the Canon cameras that I have owned/used. This includes the 1D Mark II and Mark III cameras.

My point to this thread was to compare it (when cropped from a set distance) with the Mark IV. Unfortunately I do not have one to test. From samples I have seen so far, the Mark IV has slightly lower noise but less detail when cropping a subject identically from a set distance.

The bottom line is what features do the Mark IV have that the 7D does not when photographing birds in the wild in a focal-length limited situation? Here's my list of what might be better:

-- Autofocus? Who knows. Possibly. No real comparison tests yet.
-- Faster frames per second: 8 vs 10
-- Slightly better dynamic range
-- Lower noise, but when equalizing a cropped bird from a set distance not that much greater and the 7D will out-resolve the Mark IV in this situation. It's actually close to a draw from what I have seen so far once both camera's images are processed.
-- Better build quality
-- Longer battery life, but 7D battery grip may negate this.

What else am I missing?

Please Note: This comparison is only valid in focal-length-limited situations in which you cannot walk up to the bird and take its picture. Otherwise, the Mark IV is going to provide better results.

Unfortunately 75-80% of the time that I shoot, my distance to the bird is fixed and in situations in which I can walk up to a bird to photograph it, I already have the 5D Mark II.

Sorry if I have been long-winded. I'm just trying to figure out for my needs if the Mark IV will make that much of a difference in photographing birds in the wild and I figure there is no forum better for this than this one to get other bird photographer's opinions.

Alan