Canon 7D ISO 1600 Noise
- Thread starter EMA747
- Start date
Spiritflier
Suspended / Banned
- Messages
- 2,802
- Name
- Simon (Prefer 'Si')
- Edit My Images
- No
Hi all,
I'm one of those who's going to defend the capabilities of the 7D... Not just 'cos I own one but because I sincerely believe that the quality of the images produced by this camera at high ISO's is very respectable.
The following shots were taken under the worst possible conditions... Artificial lighting (floodlights) at night with relatively fast moving subjects.
None of these shot have had ANY processing apart from the normal Adobe Camera Raw defaults and they've obviously been resized for the forum.
If you're not satisfied with that sort of quality/performance, maybe you need to change hobbies!
ISO3200
ISO2000
ISO1600
I'm one of those who's going to defend the capabilities of the 7D... Not just 'cos I own one but because I sincerely believe that the quality of the images produced by this camera at high ISO's is very respectable.
The following shots were taken under the worst possible conditions... Artificial lighting (floodlights) at night with relatively fast moving subjects.
None of these shot have had ANY processing apart from the normal Adobe Camera Raw defaults and they've obviously been resized for the forum.
If you're not satisfied with that sort of quality/performance, maybe you need to change hobbies!
ISO3200
ISO2000
ISO1600
AndyElliott
Suspended / Banned
- Messages
- 500
- Name
- Andy
- Edit My Images
- Yes
I sort of know what you mean... The ISO performance, IMHO, is fine. The resolution is a waste - few lenses can make use of it, including many of those oh-so-beloved white ones. Where the resolution can help is when you scale back the image, you sample from more pixels and the noise looks a lot better. I think it may be something to do with the colour rendition. That may be more a poor implementation in Adobe Camera Raw (which must account for a majority of images, I suspect). I ended up doing my own profiles for my E-3 because I wasn't happy with the standard ACR ones.
I don't put massive store in numbers, but the dxomark for the 7D compared with the D300 (or even my now venerable E-3, hopefully to be upgraded after Photokina) shows tonal range and colour sensitivity on the 7D a bit behind. This may well be a red-herring. I think it might be one of those things that is more noticeable in poor conditions/poor exposure - the 7D can certainly produce excellent images in the right hands.
Andy
trencheel303
Suspended / Banned
- Messages
- 4,888
- Edit My Images
- No
While I agree with this somewhat, there must only be some truth in that rule otherwise L glass wouldn't sell, no one would own an EOS 1, and instead the pros would use a canon 1000D and an 18-55.
If the gear really does make no difference at all then it doesn't explain why all the pros use the top end equipment. I think it's a combination of both that helps. I mean, a good togger can't make a 1000D do 10FPS can they.
malla1962
Suspended / Banned
- Messages
- 2,677
- Edit My Images
- No
I must admit I love my 7d for what I got it for, there is no wow factor for me with the images it produces but then again I don't think images from any camera give me the wow factor, a lot of people rave about the IQ of the 5d2 but to me they are just average images.
POAH
Suspended / Banned
- Messages
- 7,882
- Edit My Images
- No
malla1962
Suspended / Banned
- Messages
- 2,677
- Edit My Images
- No
Looks good mate.7D at 6,400 ISO : the edits....
the result....
Dogfish_magnet
Suspended / Banned
- Messages
- 4,613
- Name
- Paul
- Edit My Images
- Yes
i can just see loads of noise in everything but the bright top & shorts :shrug:
tdodd
Suspended / Banned
- Messages
- 6,481
- Name
- Tim
- Edit My Images
- No
Yep. Unfortunately, some people can't see the wood for the trees. Their loss, I think.
AndyElliott
Suspended / Banned
- Messages
- 500
- Name
- Andy
- Edit My Images
- Yes
Yep. Unfortunately, some people can't see the wood for the trees. Their loss, I think.
Non-togs are often a better judge of a good pic - they don't focus on the irrelevant... Nowt wrong with the basket-ball player. Stick it through noise-ninja and you would be hard pressed to see the noise at all and still have a good, sharp image.
You could have had a cleaner image with a 5Dmk2, but it may have had trouble focussing. You could have had the best of both worlds with a 1D or 1Ds - at a high price premium. The difference would be small that 99% of people would not give a monkeys about it...
Andy
AndyElliott
Suspended / Banned
- Messages
- 500
- Name
- Andy
- Edit My Images
- Yes
Freester
Suspended / Banned
- Messages
- 1,602
- Name
- Mark
- Edit My Images
- Yes
Dogfish_magnet
Suspended / Banned
- Messages
- 4,613
- Name
- Paul
- Edit My Images
- Yes
its not so much the high iso performance, I've seen lots of bird shots on here at relatively moderate iso's that don't look right, they appear to lose detail to noise and the backgrounds look very grainy , the image might not be any noisier than other bodies its just that the noise appears to detract more from the image from a 7D.
AndyElliott
Suspended / Banned
- Messages
- 500
- Name
- Andy
- Edit My Images
- Yes
AndyElliott
Suspended / Banned
- Messages
- 500
- Name
- Andy
- Edit My Images
- Yes
I suspect that some folk are trying to use those 18MP for excessive cropping. If the lens isn't sharp enough at a pixel-level (which most won't be at 18MP), that adds to the signal-to-noise woes, especially when you try to sharpen your way out of trouble. Treat it like a 40D (in terms of resolution) with better noise, DR, fps and AF and there will be no problem.
trencheel303
Suspended / Banned
- Messages
- 4,888
- Edit My Images
- No
Sit back, look away for a minute, and then loook at it like a normal person would. It's totally fine... thanks tdodd for posting it.
I've realised the same now for my 450D at 1600 ISO and have started to use it when I needed to. For ISO 6400 the images posted are very , very good.
tdodd
Suspended / Banned
- Messages
- 6,481
- Name
- Tim
- Edit My Images
- No
Here's a way to think about the relationship between ISO, noise and image viewing size. It might be complete nonsense, but I think it would reel in expectations to a more reasonable level, and I'm sure there is some logic that works here....
Let's suppose that as far as noise is concerned (ignore sharpness for now) a well exposed file from a 7D shot at 100 ISO can indeed be viewed at 100%, with noise at acceptable levels. On a 100 DPI monitor that would produce a virtual image of 52" across, or over 4'. That's equivalent to watching a TV with a screen diagonal of 62", which is pretty huge and not something you would usually watch from 12" away, as you might when examining an image on a monitor.
Now, if you were to raise the ISO to 200 and also double the shutter speed to maintain the same exposure the sensor would only receive half the light. The ISO has nothing to do with it, but by doubling the shutter speed the sensor is only exposed for half as long. Thus only half as many photons are received. With only half the light captured by the sensor, perhaps there is an argument that you can only produce an image of half the area in total, if you want to preserve the perceive noise levels.
So instead of continuing to display a 200 ISO file from a 7D at 100% for a 52" image, we would need to reduce the dimensions by SQRT(2), or a factor of 1.41. Thus your viewing % should be 100/1.41 = 71%. That would maintain equivalent light density within the image, and approximately equal noise levels. Even at 71% viewing you'd still have an image that was 37" across - over 3'. That's pretty huge for analysis from 12" away.
If we continue this approach as we raise the ISO in 1 stop intervals we get the following progression in order to preserve the density of light....
- At 100 ISO, view at 100% for a 52" image
- At 200 ISO, view at 71% for a 37" image (or you could go for a common figure of 66.67%)
- At 400 ISO, view at 50% for a 26" image
- At 800 ISO, view at 35% for an 18" image (or you could go for a common figure of 33.33%)
- At 1600 ISO, view at 25% for a 13" image
- At 3200 ISO, view at 18% for a 9" image (or you could go for a common figure of 16.67%)
- At 6400 ISO, view at 12.5% for a 6.5" image
Now, based on real world experience I'd actually say that table of figures is not at all far from the truth. If you were to use NR, or print, you might well squeeze a little more size out of the high ISO images. Depending on the nature of the subject, you might actually be able to go quite a bit larger. e.g. running heavy NR on a bird's feathers, small in the frame, would destroy important details. Running heavy NR on a vehicle that was large in the frame probably wouldn't.
Of course, if you were to view the image at a distance appropriate for the image size, to view the image at 100%, or 52" across, you should be standing about 5' away from the image, not 12".
It really isn't that the camera is performing poorly. It's (IMHO) far more the case that people's expectations are in need of adjustment. Some people seem to expect that they can shoot at high ISOs and still crop in at 100%, or perhaps 50%. That's really not going to work very well, especially for subjects with fine detail that is lost in the noise, like little birds filling about 5% of the frame. If you shoot with a 7D at 1600 ISO then, according to my table above, you should consider viewing the file at no more than 25%. In other words, if you want to make a web image of 800x533, expect to crop out an area of 3200x2133 and then resize that to 800x533. Do not crop out a tiddly little area, of say 1500x1000 and then expect that to look acceptable as an 800x533 web shot. It probably won't.
Returning to my basketball shot above, note that I cropped that image before converting to JPEG for final output. The cropped dimensions are 2372x3558. By consulting my table above I see that for a 6400 ISO image I should be viewing at only 12.5% which means I should only expect to produce a file of 297x445 pixels from that crop. Is it any wonder that by displaying at 533x800 it still has signs of noise? That's a display size of 22% instead of the "proposed/recommended" 12.5%.
Let's suppose that as far as noise is concerned (ignore sharpness for now) a well exposed file from a 7D shot at 100 ISO can indeed be viewed at 100%, with noise at acceptable levels. On a 100 DPI monitor that would produce a virtual image of 52" across, or over 4'. That's equivalent to watching a TV with a screen diagonal of 62", which is pretty huge and not something you would usually watch from 12" away, as you might when examining an image on a monitor.
Now, if you were to raise the ISO to 200 and also double the shutter speed to maintain the same exposure the sensor would only receive half the light. The ISO has nothing to do with it, but by doubling the shutter speed the sensor is only exposed for half as long. Thus only half as many photons are received. With only half the light captured by the sensor, perhaps there is an argument that you can only produce an image of half the area in total, if you want to preserve the perceive noise levels.
So instead of continuing to display a 200 ISO file from a 7D at 100% for a 52" image, we would need to reduce the dimensions by SQRT(2), or a factor of 1.41. Thus your viewing % should be 100/1.41 = 71%. That would maintain equivalent light density within the image, and approximately equal noise levels. Even at 71% viewing you'd still have an image that was 37" across - over 3'. That's pretty huge for analysis from 12" away.
If we continue this approach as we raise the ISO in 1 stop intervals we get the following progression in order to preserve the density of light....
- At 100 ISO, view at 100% for a 52" image
- At 200 ISO, view at 71% for a 37" image (or you could go for a common figure of 66.67%)
- At 400 ISO, view at 50% for a 26" image
- At 800 ISO, view at 35% for an 18" image (or you could go for a common figure of 33.33%)
- At 1600 ISO, view at 25% for a 13" image
- At 3200 ISO, view at 18% for a 9" image (or you could go for a common figure of 16.67%)
- At 6400 ISO, view at 12.5% for a 6.5" image
Now, based on real world experience I'd actually say that table of figures is not at all far from the truth. If you were to use NR, or print, you might well squeeze a little more size out of the high ISO images. Depending on the nature of the subject, you might actually be able to go quite a bit larger. e.g. running heavy NR on a bird's feathers, small in the frame, would destroy important details. Running heavy NR on a vehicle that was large in the frame probably wouldn't.
Of course, if you were to view the image at a distance appropriate for the image size, to view the image at 100%, or 52" across, you should be standing about 5' away from the image, not 12".
It really isn't that the camera is performing poorly. It's (IMHO) far more the case that people's expectations are in need of adjustment. Some people seem to expect that they can shoot at high ISOs and still crop in at 100%, or perhaps 50%. That's really not going to work very well, especially for subjects with fine detail that is lost in the noise, like little birds filling about 5% of the frame. If you shoot with a 7D at 1600 ISO then, according to my table above, you should consider viewing the file at no more than 25%. In other words, if you want to make a web image of 800x533, expect to crop out an area of 3200x2133 and then resize that to 800x533. Do not crop out a tiddly little area, of say 1500x1000 and then expect that to look acceptable as an 800x533 web shot. It probably won't.
Returning to my basketball shot above, note that I cropped that image before converting to JPEG for final output. The cropped dimensions are 2372x3558. By consulting my table above I see that for a 6400 ISO image I should be viewing at only 12.5% which means I should only expect to produce a file of 297x445 pixels from that crop. Is it any wonder that by displaying at 533x800 it still has signs of noise? That's a display size of 22% instead of the "proposed/recommended" 12.5%.