Reading mode: Light Dark

Evaluating D800 images shrunken to 12 or 16mp defeats the purpose.

Go to last post
DocS said:
tony field said:
Seems to me the arguments are rather tangential. However, two things seem to emerge:

1. If you are comparing the technology of similar or dissimilar cameras, the appropriate methodology is to compare pixel level sensor information. Thus, it is reasonable to compare a 12mp and a 36mp camera at the pixel level.

2. If you are comparing the ability to deliver images of similar or dissimilar cameras, the appropriate methodology is to display the images at the same visual size - if that requires up/down resolutions, so be it.

The pixel level analysis helps explain some of the thing you see in the image analysis. Conversely, the image analysis helps you interpret the findings of the pixel analysis.

IMHO, it is about that simple. Two complimentary ways of evaluation are totally reasonable.
--
tony
http://www.tphoto.ca
Click to expand...
Well said.

What I’m saying is that, in order to judge whether or not the D800 is a true improvement over the D700 - at least where its sensor (and related firmware) is concerned - and not just a full-frame sensor with more pixels shoved in, one has to compare the images at the pixel level.
Click to expand...
No. The D800 is at a disadvantage to begin with having smaller pixels. If the D800 was made with D700 tech it would look worse on pixel level, even though the sensor tech was equally good.

If you want to compare the inherent efficiency and quality of the sensor noise wise you have to compare the pictures at similar sizes to get a fair comparison.
DocS said:
Because the reality is that if all you plan on doing is printing 8.5x11 photos, then a 6mp camera will probably look about as good as any 12, 16, 22, or 36mp camera, and the comparison becomes leveled.
Click to expand...
Because of bayer trickeries a D800 file downsized to match the resolution of another camera will always look sharper and have more detail.
 
DocS said:
Is it lost on you that the D800 has more pixels per unit area on the sensor, and that this translates to taking up more pixels on your monitor? I guess you’re unaware that the D800 images will actually show up larger on your monitor, and will show more detail....almost as if the image that the D700 was photographing the subject through a magnifying glass that magnified the image 3x.
Click to expand...
No one is unaware of that.
DocS said:
So, look at it this way:

With the D700, you have 13,888 pixels per mm^2 of sensor area. With the D800, you have 41,777 pixels per mm^2 of sensor area.

Let us say you photograph an object that falls on 1mm worth of sensor area - meaning that when the D700 photographs it, 13,888 pixels will be used to capture that image, and when the D800 photographs it, 41,777 pixels will be used. I’d want to know that ‘percentage’ of noise coming from the 41k pixels of the D800 is comparable to (or better than) that of the 13k pixels of the D700.

Does that make more sense?
Click to expand...
That of course does make sense, however it's actually a perfect opposing argument to the argument you made when you started this thread. This is what I'm talking about with the contradiction. If you want to know how much noise a certain area of the D800's sensor generates compared to the same area on the D700's sensor, you would view those areas at the same size. That would require re-sizing one or the other.

However, if you wanted to know how the sensors compared at a pixel level, rather an a sensor area level, you would compare 13,888 pixels on the D700 to 13,888 pixels on the D800. To do that without variables, you would actually have to take an image of a wider scene on the D800, assuming you want a directly comparable subject. You'd have to go far enough away with the D800 that the subject is covered by the same amount of pixels on both cameras. Maybe you should set up such a test, pixel level performance on a monitor isn't of a huge concern to me, since I'm not printing 100% crops at 72 ppi.

Most standard image quality tests use the same test scene regardless of resolution since most people already understand that you'll be able to print larger at the same ppi with higher resolution cameras.

Comparing performance when shooting the same scene viewed at the same size is a lot more relevant unless someone is going to change their shooting style to match the pixel density of every new camera they buy, making sure to shoot the scene from further away or with a wider lens to make sure they put the same number of pixels as their previous camera on any given part of the subject.

If this is how you shoot I fully understand why 100% crops are important to you.
--
http://www.flickr.com/photos/mcullenphoto/
 
DocS said:
Synnberg said:
Yes, the point is that when you downsize at 36MP image, it retains much more microdetail than a 12MP one. For example,
Click to expand...
I’m sorry, but you’re wrong.

The downsized D3x image isn’t showing more detail. It’s sharper. But it’s not showing more detail. Nor could it. It is physically impossible for a 36mp image to be shrunken down to 12mp and show more detail than a 12mp camera. Why? Because when you shrink the 36mp image down, you are also eliminating 24mp worth of information and keeping only 12mp worth. Now, it’s conceivable that the software may selectively eliminate just the right things to make the image sharper than that of a 12mp sensor, but the amount of data present in both images is going to be the same, and by that virtue, the amount of detail they will show will be virtually identical....and the sample you showed me is a perfect example of this.
Click to expand...
Look at this example that I draw on frequently:

http://diglloyd.com/blog/2009/20090109_1-NikonD3x.html

The total MTF between two systems, one a 12MP native bayer sensor, and the other a 24MP (or 36MP) bayer sensor downsampled to 12MP, is different in a frequency-specific way.

The native 12MP does not entirely sample at all frequencies up to the nyquist limit for a 12MP sample space. There is a rolloff at the top due in part to a severe AA filter, and bayer artifacts. This causes the text in the sample to become fragmented. The high-resolution capture, when downsampled to 12MP, retains a bit more of the high-frequencies up to the nyquist limit of the 12MP sample space. By oversampling -- using a more gentle AA filter I might add, or none at all -- you are compensating for the limitations of the native 12MP system.

You should look for this kind of preserved high-frequency detail in places like hair, skin, branches, grass, foliage. This is a characteristic look that you should learn to recognize. And you can especially see it in faces that take up a minority of the frame. We are evolved to be naturally keen on reading facial detail as mammals.
 
DocS said:
tony field said:
DocS said:
What I’m saying is that, in order to judge whether or not the D800 is a true improvement over the D700 - at least where its sensor (and related firmware) is concerned - and not just a full-frame sensor with more pixels shoved in, one has to compare the images at the pixel level.
Click to expand...
Completely agree. IMHO, the D800 actually performs somewhat better than the D700 at the pixel level. An obvious technological advance is visible which is impressive considering our pre-disposition to thinking "bigger pixels are better".
DocS said:
Because the reality is that if all you plan on doing is printing 8.5x11 photos, then a 6mp camera will probably look about as good as any 12, 16, 22, or 36mp camera, and the comparison becomes leveled.
Click to expand...
In a general sense, I agree. None the less, when I compare (at, say ISO 200), the 8.5x11 print from my d2x with the 5d-II, there is, for a lack of a better term, more "micro contrast" or "smoothness" to the downsized images with the larger pixels count. Of course the camera with lesser pixels can approximate the camera on these prints by (very) careful use of slight sharpening. This sharpening affects the d2x images slightly more than it does the 5d-II - but the two images are, for all intent and purposes, "identical".
Click to expand...
I agree completely. A 36mp image downsized to 12mp will look sharper than a native 12mp. But it won’t show more detail.

Although...I suppose there is one way, conceivably, that it could be made to show more detail. Obviously, when you shrink a photo, you’re eliminating data. If the software performing the photo re-size were to make an effort to selectively eliminate noise data rather than subject/image data while shrinking the 36mp photo, then the shrunken photo might consist of more ‘pure’ subject/image data, whereas the native 12mp image would have both subject/image data and noise data.

So, for example, let us say that the D800 and D700 perform equally well at the pixel-level, and that their signal-to-noise ratios are identical. If the software shrinking the D800’s image down to 12mp cuts out all of the noise data and produces a more pure image, then you might have a higher percentage of subject/image data...and in-turn, a nearly-unnoticeable amount of detail above the D700’s. I’d image that this is purely a hypothetical scenario, and that software doesn’t exist to do that. I could be wrong.
Click to expand...
Remember when I told you that most shot noise was at a high frequency that would be almost entirely eliminated through downsampling, which eliminates simply all high frequencies beyond a certain point. This is one beneficial effect of supersampling.

So your hypothesis is basically right here. Except for the added fact that there is a little better MTF up near the limit in a supersampling system, as I mentioned in a response further upstream.
 
Luke Kaven said:
The native 12MP does not entirely sample at all frequencies up to the nyquist limit for a 12MP sample space.
Click to expand...
Exactly, the 12MP native image would have to be "perfect" at the pixel level to contain the same amount of detail as a higher resolution image down sampled to 12MP. Making these points to the OP seems like a waste of time though, he doesn't seem interested in letting facts have any influence on his opinions.

--
http://www.flickr.com/photos/mcullenphoto/
 
Synnberg said:
DocS said:
If I’m trying to see the smaller details that would be captured in a large print, then yes, I’m going to view it up close. That’s the whole point of having more pixels. So that you capture more detail (which inherently be viewed from up close, unless the image is blown up even more).
Click to expand...
No, that's not the point of having more pixels. I never view my prints from an inch away, nor have any of my clients. Have you sold any prints yet?
Click to expand...
No. Nor will I ever.

But we’re not talking about what you do. We’re talking about what certain cameras are capable of doing. You need to stop with this tendency to judge the D800 based on how YOU would use it, and instead, evaluate its capabilities from a more objective standpoint.
Synnberg said:
Synnberg said:
Synnberg said:
Yes, the point is that when you downsize at 36MP image, it retains much more microdetail than a 12MP one. For example,
Click to expand...
I’m sorry, but you’re wrong.

The downsized D3x image isn’t showing more detail. It’s sharper. But it’s not showing more detail. Nor could it. It is physically impossible for a 36mp image to be shrunken down to 12mp and show more detail than a 12mp camera. Why? Because when you shrink the 36mp image down, you are also eliminating 24mp worth of information and keeping only 12mp worth. Now, it’s conceivable that the software may selectively eliminate just the right things to make the image sharper than that of a 12mp sensor, but the amount of data present in both images is going to be the same, and by that virtue, the amount of detail they will show will be virtually identical....and the sample you showed me is a perfect example of this.
Click to expand...
Unless you understand the difference between detail and microdetail, there's no point in me trying to make you understand.
Click to expand...
Well pardon me. Why don’t you tell me what “microdetail” is. Comparing the two images, I can tell from the D3s the textures of the various fabrics, and every last dot I can see on the D3x’s shrunken image, I can see on the D3s’s (albeit not as sharp). I’d say that qualifies as “microdetail”. The D3x image simply looks sharper. Is there something else you’d care to say about this comparison?
Synnberg said:
Synnberg said:
Just because you print all of your photos 5x7 doesn’t mean that there’s no use to seeing how 36mp images come out on the D800. Hell, I don’t even print my photos, and I see the relevance of evaluating the pixel-level performance of the D800.
Click to expand...
Yeah, there you go. You don't even print your images and that's exactly why you're fussing over things that real photographers don't care about. The largest commercial print I've done is a couple of storeys tall. That's much more relevant in the "Real world" than your need for looking at cat pictures at 100% on a monitor. Sucks that Nikon doesn't make a camera for your specific needs and more in line with MY real world needs.
Click to expand...
Why do you keep trying to put words and ideas in my mouth? Did I not say that I already bought a D800?

I’m simply stating that if you want to evaluate whether or not the D800’s sensor is a genuine improvement over the D700’s, you have to evaluate it at the pixel level.

And one more thing - if looking at photos of my cat is what I want to do with my cameras, that is every bit as “real-world” and every bit as “relevant” as your 5-story prints. I can assure you that, to me, photos of my family, my son, and the various things I like to photograph when I’m in the mood to take photos is every bit as important to me as your subjects are to you.

That’s why Nikon sells pro cameras to recreational photographers too...and, how ‘bout that....for the same price to boot! They charge your dumb ass the same price they charge me. Looks to me like Nikon wants my business as much as they want yours.
Synnberg said:
I'll make it simple for you. There are three scenarios.

1) When you're making small prints: A higher MP image, scaled down would exhibit better DR and microdetail than a lower MP one would.
Click to expand...
Wrong on the microdetail, assuming microdetail means what I think it means.
Synnberg said:
2) When you're making very large prints: A higher MP image would let you print these at much higher DPIs and when viewed at the intended viewing distance, would appear to have far more detail than a lower MP image would.
Click to expand...
We are arguing semantics here. Higher mp means more detail to be seen. And the distance from which it can be seen (or must be seen) is determined by how it’s printed. Got it. You can drop this infantile argument because we are in complete agreement here.
Synnberg said:
3) Viewing cat pictures at 100% on a monitor: A higher MP image would urge armchair photographers to create more whining related threads than a lower MP image would.
Click to expand...
Well, it’s us “armchair photographers” that keep Canon and Nikon making better cameras for “professional” photographers like you.
Synnberg said:
Synnberg said:
Synnberg said:
Anybody with a functional brain.
Click to expand...
Oh yeah. There you go. Very strong, convincing argument. Shows me that you really know what you’re talking about.
Click to expand...
Far more than you do. As I said before, my proof is in my gallery. Where are your gallery pictures?
Click to expand...
I’ll post them when I’m good and ready to post them.
 
MTF is the concept you are looking for.

The two systems (a) native 12MP sensor, and (b) native 24 (or36) MP sensor downsampled judiciously to 12MP have a very different total MTF.

System (a) rolls off beginning at lower frequencies than system (b). You are seeing in system (b) more of the high-frequency components that system (a) cannot represent faithfully.

Whether you are also seeing some additional sharpening artifacts is somewhat debatable.
 
DocS said:
Well, it’s us “armchair photographers” that keep Canon and Nikon making better cameras for “professional” photographers like you.
Click to expand...
Actually it's the regular consumer buying point and shoot cameras, low end DSLRs, and kit lenses that give these companies their highest profit margins. Most of those people would never even spend time posting on a camera equipment forum like this.
DocS said:
I’ll post them when I’m good and ready to post them.
Click to expand...
What do you shoot with now? What are your primary subjects? You can view mine at the link in my signature. I think a lot of the images turned out ok despite the fact that you seem to think I don't even understand how viewing images on a monitor works.

I'd like to see some samples from a real expert. I'd imagine someone who spends most of their time worrying about pixel level performance must have already eliminated all the other variables that effect final image quality with absolutely perfect technique and processing.

It's hard to take your argument seriously without some example of why this is so important to you.

--
http://www.flickr.com/photos/mcullenphoto/
 
DocS said:
I’d want to know that ‘percentage’ of noise coming from the 41k pixels of the D800 is comparable to (or better than) that of the 13k pixels of the D700.

Does that make more sense?
Click to expand...
Yes it is. That is exactly the proposition. You are correct for centering on this one thing.
 
DocS said:
What I’m saying is that, in order to judge whether or not the D800 is a true improvement over the D700 - at least where its sensor (and related firmware) is concerned - and not just a full-frame sensor with more pixels shoved in, one has to compare the images at the pixel level.
Click to expand...
Remember earlier I said you aggregate pixels according to whether photons are plentiful or scarce. When photons are plentiful, you get good response -- excellent -- at the (small) pixel level. When photons are scarce, you need to aggregate them to get an acceptable signal-to-noise ratio. Having some additional information about photon location from a high-resolution sensor makes captures information-rich.
DocS said:
Because the reality is that if all you plan on doing is printing 8.5x11 photos, then a 6mp camera will probably look about as good as any 12, 16, 22, or 36mp camera, and the comparison becomes leveled.
Click to expand...
There are significant differences in the comparative "total MTF" between two kinds of capture systems The level of high-frequency information transferred \is not equal between (a) captures from a native 12/16MP sensor and (b) captures from a native 24/36MP sensor downsampled to 12/16MP. This might be partly due to a difference on optical-low pass filters, bayer demosaic artifacts, or things not yet examined.

Try it.
 
DocS said:
What I’m saying is that, in order to judge whether or not the D800 is a true improvement over the D700 - at least where its sensor (and related firmware) is concerned - and not just a full-frame sensor with more pixels shoved in, one has to compare the images at the pixel level.
Click to expand...
If you really want to compare sensor improvement between the D700 era and now you'd need either a 12MP D800 or a 36MP D700. Fact is there's no 12MP camera with D800 technique.

Your best bet is to compare a DX crop from the D800 to the D7000.

Personally I don't care too much about such a comparison. I want to know what my clients will see and they judge an image at print size.

--
Philip

 
Fearless_Photog said:
DocS said:
Is it lost on you that the D800 has more pixels per unit area on the sensor, and that this translates to taking up more pixels on your monitor? I guess you’re unaware that the D800 images will actually show up larger on your monitor, and will show more detail....almost as if the image that the D700 was photographing the subject through a magnifying glass that magnified the image 3x.
Click to expand...
No one is unaware of that.
Fearless_Photog said:
So, look at it this way:

With the D700, you have 13,888 pixels per mm^2 of sensor area. With the D800, you have 41,777 pixels per mm^2 of sensor area.

Let us say you photograph an object that falls on 1mm worth of sensor area - meaning that when the D700 photographs it, 13,888 pixels will be used to capture that image, and when the D800 photographs it, 41,777 pixels will be used. I’d want to know that ‘percentage’ of noise coming from the 41k pixels of the D800 is comparable to (or better than) that of the 13k pixels of the D700.

Does that make more sense?
Click to expand...
That of course does make sense, however it's actually a perfect opposing argument to the argument you made when you started this thread. This is what I'm talking about with the contradiction. If you want to know how much noise a certain area of the D800's sensor generates compared to the same area on the D700's sensor, you would view those areas at the same size. That would require re-sizing one or the other.
Click to expand...
Incorrect. How can you view them at the same size when the image on your monitor is 3x larger with the D800? Per mm^2 of sensor real estate, the D800 is going to use up 3x the monitor space as your D700. From there, you can compare the ‘percentage’ or ‘severity’ of noise in those unequally-sized images. In other words, the “percentage” of noise is the common feature that you can compare. But when you shrink the D800’s image down to match the D700’s size, you are deleting data and not getting an accurate representation of what the D800’s sensor is doing.
Fearless_Photog said:
However, if you wanted to know how the sensors compared at a pixel level, rather an a sensor area level, you would compare 13,888 pixels on the D700 to 13,888 pixels on the D800.
Click to expand...
That too would be perfectly acceptable, as it would achieve the same goal as comparing 1mm^2 worth of pixels on both sensors (41k vs. 13k pixel images), as you are not re-sizing anything. It’s the re-sizing that renders the comparison inaccurate. Cropping is a-ok.

To do that without variables, you would actually have to take an image of a wider scene on the D800, assuming you want a directly comparable subject. You'd have to go far enough away with the D800 that the subject is covered by the same amount of pixels on both cameras. Maybe you should set up such a test, pixel level performance on a monitor isn't of a huge concern to me, since I'm not printing 100% crops at 72 ppi.
Fearless_Photog said:
Most standard image quality tests use the same test scene regardless of resolution since most people already understand that you'll be able to print larger at the same ppi with higher resolution cameras.

Comparing performance when shooting the same scene viewed at the same size is a lot more relevant unless someone is going to change their shooting style to match the pixel density of every new camera they buy, making sure to shoot the scene from further away or with a wider lens to make sure they put the same number of pixels as their previous camera on any given part of the subject.
Click to expand...
Having higher resolution (with an equivalent or better signal-to-noise ratio) is in some ways like having lower resolution camera fitted with a lens with more zoom (e.g. longer focal length). Some of us will take that into account when shooting, knowing that the D800 will capture additional details that we can crop around.
Fearless_Photog said:
If this is how you shoot I fully understand why 100% crops are important to you.
Click to expand...
Well, I think it becomes important to everyone at some point. I think the D800 and its huge resolution will get some people to depend on the added resolution from time to time as a form of zoom.
Fearless_Photog said:
http://www.flickr.com/photos/mcullenphoto/
Click to expand...
 
Luke Kaven said:
DocS said:
tony field said:
DocS said:
What I’m saying is that, in order to judge whether or not the D800 is a true improvement over the D700 - at least where its sensor (and related firmware) is concerned - and not just a full-frame sensor with more pixels shoved in, one has to compare the images at the pixel level.
Click to expand...
Completely agree. IMHO, the D800 actually performs somewhat better than the D700 at the pixel level. An obvious technological advance is visible which is impressive considering our pre-disposition to thinking "bigger pixels are better".
Luke Kaven said:
Because the reality is that if all you plan on doing is printing 8.5x11 photos, then a 6mp camera will probably look about as good as any 12, 16, 22, or 36mp camera, and the comparison becomes leveled.
Click to expand...
In a general sense, I agree. None the less, when I compare (at, say ISO 200), the 8.5x11 print from my d2x with the 5d-II, there is, for a lack of a better term, more "micro contrast" or "smoothness" to the downsized images with the larger pixels count. Of course the camera with lesser pixels can approximate the camera on these prints by (very) careful use of slight sharpening. This sharpening affects the d2x images slightly more than it does the 5d-II - but the two images are, for all intent and purposes, "identical".
Click to expand...
I agree completely. A 36mp image downsized to 12mp will look sharper than a native 12mp. But it won’t show more detail.

Although...I suppose there is one way, conceivably, that it could be made to show more detail. Obviously, when you shrink a photo, you’re eliminating data. If the software performing the photo re-size were to make an effort to selectively eliminate noise data rather than subject/image data while shrinking the 36mp photo, then the shrunken photo might consist of more ‘pure’ subject/image data, whereas the native 12mp image would have both subject/image data and noise data.

So, for example, let us say that the D800 and D700 perform equally well at the pixel-level, and that their signal-to-noise ratios are identical. If the software shrinking the D800’s image down to 12mp cuts out all of the noise data and produces a more pure image, then you might have a higher percentage of subject/image data...and in-turn, a nearly-unnoticeable amount of detail above the D700’s. I’d image that this is purely a hypothetical scenario, and that software doesn’t exist to do that. I could be wrong.
Click to expand...
Remember when I told you that most shot noise was at a high frequency that would be almost entirely eliminated through downsampling, which eliminates simply all high frequencies beyond a certain point. This is one beneficial effect of supersampling.

So your hypothesis is basically right here. Except for the added fact that there is a little better MTF up near the limit in a supersampling system, as I mentioned in a response further upstream.
Click to expand...
Now when you say most shot noise occurs “at high frequency”, does that means that noise shows up as blues/violets on the image? Something else, perhaps?
 
DocS said:
Fearless_Photog said:
DocS said:
Is it lost on you that the D800 has more pixels per unit area on the sensor, and that this translates to taking up more pixels on your monitor? I guess you’re unaware that the D800 images will actually show up larger on your monitor, and will show more detail....almost as if the image that the D700 was photographing the subject through a magnifying glass that magnified the image 3x.
Click to expand...
No one is unaware of that.
DocS said:
So, look at it this way:

With the D700, you have 13,888 pixels per mm^2 of sensor area. With the D800, you have 41,777 pixels per mm^2 of sensor area.

Let us say you photograph an object that falls on 1mm worth of sensor area - meaning that when the D700 photographs it, 13,888 pixels will be used to capture that image, and when the D800 photographs it, 41,777 pixels will be used. I’d want to know that ‘percentage’ of noise coming from the 41k pixels of the D800 is comparable to (or better than) that of the 13k pixels of the D700.

Does that make more sense?
Click to expand...
That of course does make sense, however it's actually a perfect opposing argument to the argument you made when you started this thread. This is what I'm talking about with the contradiction. If you want to know how much noise a certain area of the D800's sensor generates compared to the same area on the D700's sensor, you would view those areas at the same size. That would require re-sizing one or the other.
Click to expand...
Incorrect. How can you view them at the same size when the image on your monitor is 3x larger with the D800? Per mm^2 of sensor real estate, the D800 is going to use up 3x the monitor space as your D700. From there, you can compare the ‘percentage’ or ‘severity’ of noise in those unequally-sized images. In other words, the “percentage” of noise is the common feature that you can compare. But when you shrink the D800’s image down to match the D700’s size, you are deleting data and not getting an accurate representation of what the D800’s sensor is doing.
Click to expand...
You take into account not just the physical size of the displayed (or printed) image, but also viewing distance, and therefore apparent size , which is the relevant perceptual metric in the end.
DocS said:
Having higher resolution (with an equivalent or better signal-to-noise ratio) is in some ways like having lower resolution camera fitted with a lens with more zoom (e.g. longer focal length). Some of us will take that into account when shooting, knowing that the D800 will capture additional details that we can crop around.
Click to expand...
You can always use the DX crop, for example, in which case the camera will perform with total noise equivalent to a DX camera.
DocS said:
DocS said:
If this is how you shoot I fully understand why 100% crops are important to you.
Click to expand...
Well, I think it becomes important to everyone at some point. I think the D800 and its huge resolution will get some people to depend on the added resolution from time to time as a form of zoom.
Click to expand...
When photons are plentiful, you can collect enough of them to have 36M good pixels. When photons are scarce, you will tend to aggregate pixels and trade off resolution to gain quality.
 
DocS said:
jfriend00 said:
When you've offered a fair way to compare the noise between different resolution cameras, then you can dump on the methods being used. It's a fact that it's not a legit comparison to compare pixel level noise in two cameras that has massively different resolution. You have to compare the two at some common level (e.g. a web size image or a 6 foot high print or something in between). But, you can't look at 100% pixel level noise at 36MP and compare that to 100% pixel level noise at 12MP and draw any meaningful conclusions from that comparison.
Click to expand...
Oh?

How would comparing D700 and D800 images at their native resolutions NOT tell you anything meaningful? It might tell you that the D800, with its higher resolution, produces more noise per unit-area of sensor, and that the only way the image quality becomes comparable to the D700’s is is by shrinking it down (which gets rid of much of the noise).
Click to expand...
If you compare the D700 and D800 at their native resolutions, you are measuring the ability of one pixel on the D800 vs. one pixel on the D700. That is one measure of technology and if the D800 beats the D700, then hands-down, the D800 wins in all regards.

But, life isn't quite that simple because if the D800 produces more single pixel noise than a D700, that does not mean that it doesn't produce a much better image than the D700 when compared more equally. Give the two cameras the same lens, the same scene, the same light and ask them to produce the same output (same large print, same cropped web image, same poster, etc...) and the D800 likely will produce a much better image with both more detail and lower visible noise. That's because there is a lot more that goes into making of an actual image than comparing single pixels from two drastically different resolution sensors.

The very fact that the D800 has many more pixels representing the same piece of the image means that it's per pixel noise doesn't count as much in the image. The effect of the noise from a given pixel in the D800 is smaller than that of the D800 because there are so many more D800 pixels in a given piece of image. The D800 pixel nose will affect areas of fine detail less.

Thus, if you're trying to measure the camera's ability to produce an actual image (print or screen), you have to compare two actual results side by side.

Said another way, comparing a 100% view of a D800 image and a 100% view of a D700 image is like using a magnifying glass on one image and not the other. Yes, you can see some things about that image with the magnifying glass and perhaps learn some things about the technology behind it, but it isn't a fair comparison of what kind of image the two can produce to use a magnifying glass on only one of the two images.

Here's another way to look at it. Imagine that the D800 per pixel noise was 1.5x worse than that of the D700 per pixel noise. Would you immediately conclude that the D800 produced much noisier images than the D700 and isn't as good in low light/high ISO? Your argument makes it sound like you would. If you did, you'd be missing what actually counts in the end which is resolvable detail in a final image at a particular size (and other elements of IQ). It is possible for a higher resolution camera with more per pixel noise to produce a high quality image than a lower resolution camera with less per pixel noise. Thus, the per pixel noise is not the only relevant factor here. Thus, you can't compare 100% views of two drastically different images and expect to be judging the ability of these two cameras to produce a quality image when given the same conditions.

Now, I will give you that looking at the two 100% views does tell you something about the new technology. It just doesn't tell you the whole story and even if the D800's per pixel noise is a little worse than the D700's, it may very well produce much higher IQ images because it has 3x the number of pixels to contribute detail and finer grained noise and more averaging of the random noise.
--
John
Gallery: http://jfriend.smugmug.com
Popular: http://jfriend.smugmug.com/popular
Portfolio: http://jfriend.smugmug.com/portfolio
 
I am aware of that and I do shoot a lot more JPEG than RAW these days, but I am talking about a much more evolved system for this that makes the camera far more versatile. For example, sell us a D5 with a battery grip system that is robust and just nix the D900 as the smaller stepchild to the big fat body. Then implement the nifty resolution versus noise reducing scaling and we are set.

Nikon has always had a smaller version of the pro body in their line up that invariably pairs nicely with the larger sibling with pros often preferring it to the larger body.

But the way we deal with ISO, noise reduction and resolution is now so dated it is not even funny, and honestly, I think big camera makers like Nikon and Canon are talking us for a financial ride because of it...
powertoold said:
You can have the D800 scale your 36mp images to 20mp or less, but the output is JPG only.
Click to expand...
--

'Digital is like shaved legs on a man - very smooth and clean but there is something
acutely disconcerting about it.'
 
When you look at an image, would you really care how many MP it had from the beginning!?
Of course not, no one cares! It is nothing but plain silly .

What you would care about, is what one looks better.

A photographer would compare prints. A computer guy might use a monitor. No one but a fool would use different sized images.
 
Ola Forsslund said:
When you look at an image, would you really care how many MP it had from the beginning!?
Of course not, no one cares! It is nothing but plain silly .

What you would care about, is what one looks better.

A photographer would compare prints. A computer guy might use a monitor. No one but a fool would use different sized images.
Click to expand...
You would get a better 12MP final from a 36MP capture downsampled to 12MP than you would from a native 12MP capture. Other than that, you would not care how your image started in life.
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top