Grainy images from my 5D MK IV?

[Morlin]

Hi.

Not sure if it´s just me or so but I feel that the images from my 5D MK IV are more grainy than from my 5d MKIII. Can that be correct? Can it be just a feeling I get because of the larger files that makes it possible to get closer to the fine details or can there be a quality issue or just how the files are? I also feel that they are softer than from my 5D MKIII.

Please give me your thoughts on this.

Regards / Johan

 

[AdamBotond]

I'm not a user of any of those two. You may feel its more grainy due to MP increase over the prodecessor if you look pics at 100%. In fact however, 5D IV should be superior to 5D III by a stop or so at any given ISO. Downsampling the 5D IV images to 22 MP would make a fair comparison between the two cameras and the successor should look far better especially at mid and high iso.

My two cents only.

 

[LDS]

I'm not a user of any of those two. You may feel its more grainy due to MP increase over the prodecessor if you look pics at 100%. In fact however, 5D IV should be superior to 5D III by a stop or so at any given ISO. Downsampling the 5D IV images to 22 MP would make a fair comparison between the two cameras and the successor should look far better especially at mid and high iso.

AFAIK actually you have to compare noise at 100%. Downsampling would reduce noise.

 

[AdamBotond]

Downsampling would reduce noise.

Exactly. So why not compare the high iso capabilities of two cameras with different MP count at the same resolution to make the comparison independent from MP count?

 

[YuengLinger]

I think grain doex show a bit more whe images are underexposed, especially overcast skies.

Otherwise, I'm getting cleaner images up to about ISO 5000. From there the same.

 

[LDS]

Exactly. So why not compare the high iso capabilities of two cameras with different MP count at the same resolution to make the comparison independent from MP count?

If you downsample, you're being "artificially" removing noise from one image and not the other, due to the effects of the downsampling algorithm. So you're not comparing the sensor ISO capabilities in an "independent" way.

If you want a simpler comparison, you should shoot the same subject so it has the same size in pixels in the images, then you can compare both at 1:1 and see which one has more noise without applying any transformation to the image.

Then downsampling a large image will usually help in reducing noise also.

 

[YuengLinger]

Exactly. So why not compare the high iso capabilities of two cameras with different MP count at the same resolution to make the comparison independent from MP count?

If you downsample, you're being "artificially" removing noise from one image and not the other, due to the effects of the downsampling algorithm. So you're not comparing the sensor ISO capabilities in an "independent" way.

If you want a simpler comparison, you should shoot the same subject so it has the same size in pixels in the images, then you can compare both at 1:1 and see which one has more noise without applying any transformation to the image.

Then downsampling a large image will usually help in reducing noise also.

This might be great for benchmarks in a lab, but a photographer who knows his/her equipment, has good vision, and decent software can compare shots from different cameras and decide IQ. So for real life, downsizing, etc really isn't necessary.

What is more important for an individual to judge differences is making sure that exposure is near optimal in images compared.

I'm just saying we can judge very, very well without having to count pixels, downsize, etc. Load a RAW from, say a G12 and a 5DIII, review correctly exposed images at higher ISO's, should be obvious. You don't need to down sample, etc, and if it is your money, you don't need your own perception, opinion verified by a NASA lab.

I did it with my 5DIV and 5DIII and can see that, as stated earlier, the 5DIV is slightly better up to about ISO 5000--TO MY EYES. And my eyes are the ones which count because it's my money.

But, again as stated earlier, the 5DIII appears to be slightly less grainy when an image is underexposed.

Another difference--the 5DIV images are definitely responding better to sharpening and noise reduction.

Overall, with slightly better IQ and close to drastically better AF, a very worthy upgrade for me. Plus all the nice little extra features added to the viewfinder and menus and controls.

I'm NOT dismissing the scientific, objective approach! But there are several respected websites that offer 100% comparisons with megapixel counts considered. Those are useful for helping me decide what to buy. But until I see with my own eyes, I really can't judge what suits my own tastes/needs/preferences/requirements...

For example, from the sites I saw and the specs, the 80D seemed an easy winner. But I bought one which had AF problems, then another that was working great. In both cases I was disappointed with the IQ at ISO 640 and over. Even ISO 400 had a kind of grain/noise that I found unappealing.

Mentioned the 80D so it is clear I'm not a fanboy shouting rah, rah, rah for the 5DIV. When they get it right, great, when not, too bad.

 

[privatebydesign]

Noise is collected, like light, by area. So to make a fair comparison you need to compare per sensor area. You wouldn't compare two cars with different top speeds at those different speeds to work out a mpg figure would you?

So the only fair way to compare output is to look at same sized output of the same area of the image, not pixel numbers (if the pixel numbers are different) nor zoom percentages. So, look at full screen for both (or the same subject magnification/size on screen), or a same sized print of the same area of the frame, or down sample the higher resolution to the lower resolution to compare noise, or up sample the lower resolution to the higher resolution to compare detail and resolving power.

Now there is an easy non destructive way to compare same size on screen when looking at different pixel density. In PS go to Image: Image Size. from there select a size that will fit on half your screen 8x12 or something and make sure resample is not selected. Do this to both images. The go Window: Arrange: 2 Up Horizontally (or Vertically depending on image orientation). Then go View: Print Size in both windows. Both images will display the same and if you want to scroll both equally just click H then use the shift tab with the mouse and you are good.

Or print out two test images to the same size with the same area of the image from both cameras.

At same % magnification views the higher density pixel sensor is at a disadvantage because you are magnifying it more, the comparison is not valid.

 

[YuengLinger]

Noise is collected, like light, by area. So to make a fair comparison you need to compare per sensor area. You wouldn't compare two cars with different top speeds at those different speeds to work out a mpg figure would you?

So the only fair way to compare output is to look at same sized output of the same area of the image, not pixel numbers (if the pixel numbers are different) nor zoom percentages. So, look at full screen for both (or the same subject magnification/size on screen), or a same sized print of the same area of the frame, or down sample the higher resolution to the lower resolution to compare noise, or up sample the lower resolution to the higher resolution to compare detail and resolving power.

Now there is an easy non destructive way to compare same size on screen when looking at different pixel density. In PS go to Image: Image Size. from there select a size that will fit on half your screen 8x12 or something and make sure resample is not selected. Do this to both images. The go Window: Arrange: 2 Up Horizontally (or Vertically depending on image orientation). Then go View: Print Size in both windows. Both images will display the same and if you want to scroll both equally just click H then use the shift tab with the mouse and you are good.

Or print out two test images to the same size with the same area of the image from both cameras.

At same % magnification views the higher density pixel sensor is at a disadvantage because you are magnifying it more, the comparison is not valid.

This makes a lot of sense! Will try.

 

[StudentOfLight]

What is your post-production workflow? Could you share a screenshot or 100% crop to demonstrate your issue?

 

[YuengLinger]

Do a 100% view of a 10MP G12, then 100% of 5D3 or 5D4. ISO 800. Which llooks noisier?

Doesn't matter what's ''fair.''

When we sharpen or do noise reduction, we do so at 100%. And that's where we judge IQ.

 

[privatebydesign]

Do a 100% view of a 10MP G12, then 100% of 5D3 or 5D4. ISO 800. Which llooks noisier?

Doesn't matter what's ''fair.''

When we sharpen or do noise reduction, we do so at 100%. And that's where we judge IQ.

Don't agree.

I judge IQ at the image level, it's the 'image' part of 'image quality'. Pixel level is irrelevant because I am not outputting individual pixels, or even a screens worth of them (4-12 million), I am outputting 20-30-50 million of them into an image.

100% views are entirely fallacious unless you are so focal length limited you are outputting at 1:1, whatever the sensor size.

But you are conflating several issues. The OP speculates the 5D MkIV is 'noisier at 100% view' than the MkIII, my point is that of course it is! Because it isn't a fair/relevant comparison, look at the 5D MkIII and 5D MkIV at same sized output (an image) and the 5D MkIV is less noisy and has more detail. It is relevant to appreciate that pixel density/pixel size is relevant to best practice processing, sharpen and noise reduction at 100% (sometimes), i.e. you process a MkIII file and a MkIV file differently, doesn't impact the fact that at a fair/relevant comparison (same sized output) the MkIV has higher IQ than the MkIV in noise and detail.

The fact that smaller sensors, or a smaller sensor area, is noisier is moot and entirely irrelevant.

The fact that smaller pixels are noisier is moot and entirely irrelevant.

To be sure, you can compare a banana to a camel. But if you want to know what is going to give you better IQ then for same sized sensors you need to compare same area. For different sized sensors you need to compare entire sensor area. Preferably you will make those comparisons at similar sizes to your personal output sizes.

To put that in perspective, I have a 27" Apple monitor, it has a pixel density of 109PPI. If I look at a 5D MkIII file at 100% it is the equivalent of looking at a 53" x 35" print from 20", if I do the same with a 5D MkIV file I would be looking at the equivalent of a 62" x 41" print 20" away! That is 40% more area. As an image how often do you make those sized outputs? Again, 100% view is entirely spurious unless you are outputting at 1:1 and the only relevant IQ comparison between same sized sensors with different pixel densities is same sensor area comparisons.

 

[YuengLinger]

Do a 100% view of a 10MP G12, then 100% of 5D3 or 5D4. ISO 800. Which llooks noisier?

Doesn't matter what's ''fair.''

When we sharpen or do noise reduction, we do so at 100%. And that's where we judge IQ.

Don't agree.

I judge IQ at the image level, it's the 'image' part of 'image quality'. Pixel level is irrelevant because I am not outputting individual pixels, or even a screens worth of them (4-12 million), I am outputting 20-30-50 million of them into an image.

100% views are entirely fallacious unless you are so focal length limited you are outputting at 1:1, whatever the sensor size.

But you are conflating several issues. The OP speculates the 5D MkIV is 'noisier at 100% view' than the MkIII, my point is that of course it is! Because it isn't a fair/relevant comparison, look at the 5D MkIII and 5D MkIV at same sized output (an image) and the 5D MkIV is less noisy and has more detail. It is relevant to appreciate that pixel density/pixel size is relevant to best practice processing, sharpen and noise reduction at 100% (sometimes), i.e. you process a MkIII file and a MkIV file differently, doesn't impact the fact that at a fair/relevant comparison (same sized output) the MkIV has higher IQ than the MkIV in noise and detail.

The fact that smaller sensors, or a smaller sensor area, is noisier is moot and entirely irrelevant.

The fact that smaller pixels are noisier is moot and entirely irrelevant.

To be sure, you can compare a banana to a camel. But if you want to know what is going to give you better IQ then for same sized sensors you need to compare same area. For different sized sensors you need to compare entire sensor area. Preferably you will make those comparisons at similar sizes to your personal output sizes.

To put that in perspective, I have a 27" Apple monitor, it has a pixel density of 109PPI. If I look at a 5D MkIII file at 100% it is the equivalent of looking at a 53" x 35" print from 20", if I do the same with a 5D MkIV file I would be looking at the equivalent of a 62" x 41" print 20" away! That is 40% more area. As an image how often do you make those sized outputs? Again, 100% view is entirely spurious unless you are outputting at 1:1 and the only relevant IQ comparison between same sized sensors with different pixel densities is same sensor area comparisons.

Apart from the banana and camel analogy, which went over my head, I can understand what you are saying about pixel density and noise.

But what you might be not able to see is all the attempts to make a "fair" comparison are irrelevant. When I look at my wife's G12 at 100% at ISO 400 and up, I feel much better about spending the money to buy a FF camera.

My point, again, is that photographers by in large make judgments about IQ when working on an image, and we've been taught by Adobe and many pundits that sharpening and noise reduction should be done at 100%. So if you zoom in and see, gasp, a ton of noise on a new camera compared to your old camera, theory goes out the window and you wonder what you just paid for.

And if this is a "wrong" way to judge, all the better, because it has forced the industry to keep making better sensors that are cleaner at higher ISO's.

 

[StudentOfLight]

If there is no graininess at pixel level then there will be no graininess at image level. For example, in Lightroom (with appropriate "detail" settings) you can avoid sharpening image noise which is quite often the cause of images looking excessively grainy.

Look how well the grain is controlled with this image taken with the EOS 80D at ISO 16,000: http://www.canonrumors.com/forum/index.php?topic=31329.msg639212#msg639212
I have no doubt that the 5D-IV can perform similarly if not better as it is a new-generation sensor with a considerably larger pixel pitch than the 80D.

 

[Mikehit]

When the 50D came out, there were exactly the same arguments.
The 40D was 10MP the 50D was 15MP and this was really the start of the megapixel wars when many considered 12MP more than anyone would ever need (anyone recall Nikon saying that as a corporate statement.....until they brought out the D800?). People kept on saying 'look at them on the computer, the 50D is noisier'. And it was because the automatic thing back then was to look at them 100% on the screen and people were looking at the noise per pixel. Fools.

People who made images that had everyone saying 'WOW!' were saying 'I don't care about pixels, all I know is that the 50D makes more detailed images with noise being less visible, and with more dynamic range'.
These were the people who were creating an image and it was the image (not the pixels) that were impressing everyone.

Of the reviews I have seen, those who have done an image-based review (not a pixel-based review) has said that the 5DIV is better than the 5D3 by a stop or so. Notably the ones doing image-based reviews are working professionals whose livelihood depends on it.
Guess who I am more impressed by?

 

[privatebydesign]

Apart from the banana and camel analogy, which went over my head, I can understand what you are saying about pixel density and noise.

But what you might be not able to see is all the attempts to make a "fair" comparison are irrelevant. When I look at my wife's G12 at 100% at ISO 400 and up, I feel much better about spending the money to buy a FF camera.

My point, again, is that photographers by in large make judgments about IQ when working on an image, and we've been taught by Adobe and many pundits that sharpening and noise reduction should be done at 100%. So if you zoom in and see, gasp, a ton of noise on a new camera compared to your old camera, theory goes out the window and you wonder what you just paid for.

And if this is a "wrong" way to judge, all the better, because it has forced the industry to keep making better sensors that are cleaner at higher ISO's.

No you are missing my point. The only people who have an interest in 100% views are severely focal length limited shooters who are subsequently limited to very modest electronic output sizes, measurebators, fools, and people who have bought in to the ad mans crap. What difference does it make if you feel better looking at your wives files at 100% when you don't output that? You output an image.

Back in 2009 Michael Reichmman did a comparison between a then top of the range 39MP medium format Hasselblad and a Canon G10 P&S. In good light at equivalent settings nobody could reliably tell the difference in 13" x 19" prints. The size of the sensor nor the pixels was relevant, the comparison was same sized output. That is all that is ever relevant, same sized output.

Now if you want to make bigger prints, crop harder, do action shooting, high fps, use faster lenses or just have a wider lens choice, shoot in worse light, need fast AF etc etc then camera choice becomes much more nuanced. But, if you simply want to make a valid comparison between 5D MkIII and 5D MkIV "image quality" the only relevant comparison is a same sized image.

 

[Jack Douglas]

Scott, I see this and agree. However, what I perceive and even dislike in myself is the tendency to want to put every shot under a microscope. For example take a bird portrait and examine the eyelashes at maybe 10 time the natural size. I'm willing to bet a lot of folk on CR are, when the files are posted large enough, doing exactly that. Likewise on landscape shots, today we expect to be able to zoom in from a mile away to see detail and this has nothing to do with print sizes that will be output. My observation is that the public will always buy into the more megapixels is advantageous hype not unlike the small cars that always creep up in size and power. Much of this is driven by bragging rights.

Jack

 

[Mikehit]

But what you might be not able to see is all the attempts to make a "fair" comparison are irrelevant. When I look at my wife's G12 at 100% at ISO 400 and up, I feel much better about spending the money to buy a FF camera.

My point, again, is that photographers by in large make judgments about IQ when working on an image, and we've been taught by Adobe and many pundits that sharpening and noise reduction should be done at 100%. So if you zoom in and see, gasp, a ton of noise on a new camera compared to your old camera, theory goes out the window and you wonder what you just paid for.

And if this is a "wrong" way to judge, all the better, because it has forced the industry to keep making better sensors that are cleaner at higher ISO's.

The 100% viewing mantra was first spouted back when cameras were 8MP or 10MP, not the 30MP behemoths we have nowadays. A 5DIV image is 5760 MP across. On a screen 1920-pixels wide that is 3 screens wide and you are viewing it from no more than 2 feet away.
I have read comments by some professional retouchers who say that viewing 100% is not only no longer necessary but can be very misleading, and they will view images from modern cameras at 50%. They may use 100% to define the amount of sharpening they need, or for very fine selections, but for actually assessing overall image quality they go no higher than 50% because image quality is not just about sharpness. I have found out myself that noise and detail are not mutually exclusive.

 

[LDS]

No you are missing my point. The only people who have an interest in 100% views are severely focal length limited shooters who are subsequently limited to very modest electronic output sizes, measurebators, fools, and people who have bought in to the ad mans crap. What difference does it make if you feel better looking at your wives files at 100% when you don't output that? You output an image.

You're wrong, and there's plenty of literature to show it, but given the tone of this conversation I think it's useless to even try. An insult competition is not interesting.

 

[Mikehit]

Exactly. So why not compare the high iso capabilities of two cameras with different MP count at the same resolution to make the comparison independent from MP count?

If you downsample, you're being "artificially" removing noise from one image and not the other, due to the effects of the downsampling algorithm. So you're not comparing the sensor ISO capabilities in an "independent" way.

If you want a simpler comparison, you should shoot the same subject so it has the same size in pixels in the images, then you can compare both at 1:1 and see which one has more noise without applying any transformation to the image.

Then downsampling a large image will usually help in reducing noise also.

But that isn't what you do, is it?
When you are taking an image of a landscape, you take it from a specific location because it gives a specific perspective on the landscape, or your options to move around are limited. When I am thinking about moving from 5D3 to 5DIV I am not thinking 'how can I get the same number of pixels on the subject' I am thinking 'which will look better when viewed onscreen or in print' and 'will the higher pixel density be offset by pixel quality and noise'.

When you take pictures of wildlife your options on shooting are limited by the gear you have and how close you can get. If I want a picture of an eagle on its nest and I have a 600mm lens, my question is 'will the 5DIV image look better than the 5D3 image' not 'I have to use a shorter lens to get the same number of pixels on it'.

The next question is 'do these differences hold up at all ISOs'

Understanding pixel noise is important. But that is only one step in deciding if it is a better tool for taking the images you want.

Question: The ISO capabilities is only relevant in terms of what it does for the image. So if the 5DIV proves to have more noise at the pixel level than the 5D3 and everyone says that the 5DIV takes better images, do you say the sensor is worse or better?