5D III high ISO samples

[awinphoto]

ZOMG, the DPR has a set of 5D3 low light high ISO shots from a pre-prod. Given the usual DPR testing "issues", I think ISO 12800 is very usable on this cam. 8)

I have seen DPR samples, and have the same positive opinion as yours. The detail is lower @ 12800, which it is expected but still good enough for that ISO. Those images represent a smaller print since they are not full size, but overall I believe 12800 is good for smaller prints. I am pretty sure I won't need to use ISO 12800 that much, so I am happy with these results.

Not to date myself too much who was raised/trained and shot professionally with film, The ISO 25600 shots are cleaner than the old 1600 ISO film... the old film, at high ISO was A) very expensive, and B) almost unusable for anything bigger than a 4x6, and even that look like it was sneezed on. Technology with every new camera never ceases to amaze me.

 

[torger]

Canon has done some real work on the blacks, I can move the black slider in LR4 all the way to the right without seeing any banding.

Yes that was about the first thing I tested too. Banding may be gone! I'm kind of shocked, in a positive way. I've been quite pessimistic about canon noise performance, but perhaps this sensor really can put up a match against Sony Exmor! We'll see when RAW samples become available. But it looks promising.

 

[jrista]

Why the crappy post-processing? They all look very polished and no crispeness at all. Same with the 1d X samples. Who does post on these??

The camera. Canon tends to post unprocessed JPEG's strait out of the camera, and it looks like NR is cranked up as high as it will go on the high ISO shots. Nikon, on the other hand, seems to use manually processed JPEG's created from RAW, so they tend to look better than Canon samples. I think Canon is trying to be transparent about their camera's capabilities, which is something I'm appreciative of...but it makes it tough to really compare anything between brands when they do different things. The only way we are going to get a realistic comparison is to wait for someone to get their hands on all the competing bodies, take a shot of the same scene with each one, process each in the same way, and upload a visual comparison.

 

[TAR]

anyone notice the black borders or circles on the stars ..its present on some big stars not in all of them...

 

[pedro]

I just downloaded the high ISO images from DPR. All the images were taken at 3EV, so they are real low light images.

On my screen, I could see the noise at ISO 102400, but the 8.5 X 11 prints were sharp and clean. I then printed 3 more, 51200, 25,600, 12800, and they also suprised me. I could show any of these prints to someone, and they would not suspect that they were high ISO unless they had a magnifier.

These were jpegs with no additional processing by me, I would expect RAW to be better.

I certainly plan to use the ISO 25600 freely whenever I absolutely need it in low light for a fast shutter speed.

Canon has done some real work on the blacks, I can move the black slider in LR4 all the way to the right without seeing any banding.

yes they look more then okay!

http://www.dpreview.com/galleries/reviewsamples/albums/canon-eos-5d-mark-iii-low-light-iso-samples

thanks for sharing!

 

[jrista]

anyone notice the black borders or circles on the stars ..its present on some big stars not in all of them...

Looks like oversharpening halos.

 

[BL]

anyone notice the black borders or circles on the stars ..its present on some big stars not in all of them...

facepalm :-[

i noticed the same thing just moments ago in the aurora samples...

 

[WarStreet]

I just downloaded the high ISO images from DPR. All the images were taken at 3EV, so they are real low light images.

On my screen, I could see the noise at ISO 102400, but the 8.5 X 11 prints were sharp and clean. I then printed 3 more, 51200, 25,600, 12800, and they also suprised me. I could show any of these prints to someone, and they would not suspect that they were high ISO unless they had a magnifier.

These were jpegs with no additional processing by me, I would expect RAW to be better.

I certainly plan to use the ISO 25600 freely whenever I absolutely need it in low light for a fast shutter speed.

Canon has done some real work on the blacks, I can move the black slider in LR4 all the way to the right without seeing any banding.

Thanks for the test, this is great news. Unfortunately my printer died, and I really wish I could see the results myself. I guess I need to get a new printer before the 5DIII

 

[V8Beast]

Not to date myself too much who was raised/trained and shot professionally with film, The ISO 25600 shots are cleaner than the old 1600 ISO film... the old film, at high ISO was A) very expensive, and B) almost unusable for anything bigger than a 4x6, and even that look like it was sneezed on. Technology with every new camera never ceases to amaze me.

Werd. The 5DIII's ISO 1600 samples are FAR cleaner than 200 speed film. That kind of progress just boggles my mind. In fact, I'd say the ISO 1600 shots compare favorably to ISO 200-400 on my 5DC despite the fact that the MKIII has twice the resolution. I'm equally impressed with the dynamic range as well, especially in that image of the glaciers.

I was going to test the 5DIII side by side with the D800 once I got my hands on them, but with results this good, I might just cancel my D800 pre-order outright.

 

[DavidRiesenberg]

Not only the shots look great noise wise but the noise itself seems somewhat more natural and less blocky/digital to me.

 

[V8Beast]

I just downloaded the high ISO images from DPR. All the images were taken at 3EV, so they are real low light images.

On my screen, I could see the noise at ISO 102400, but the 8.5 X 11 prints were sharp and clean. I then printed 3 more, 51200, 25,600, 12800, and they also suprised me. I could show any of these prints to someone, and they would not suspect that they were high ISO unless they had a magnifier.

These were jpegs with no additional processing by me, I would expect RAW to be better.

I certainly plan to use the ISO 25600 freely whenever I absolutely need it in low light for a fast shutter speed.

Canon has done some real work on the blacks, I can move the black slider in LR4 all the way to the right without seeing any banding.

No kidding. I think this is the test you're referring to:

http://www.dpreview.com/galleries/reviewsamples/albums/canon-eos-5d-mark-iii-low-light-iso-samples

I'm trying to be as unbiased as I can, but I had a very hard time picking up on any noticeable noise up to ISO 25,600. Only at 102,400 does it become objectionable. I'm starting to suspect that my eyes are bad or my monitor is broken Like you said, the noise in the blacks is almost non-existent. It looks like the native ISO range is a legit representation of the MKIII's true low-light capabilities.

Kudos to Canon for creating a high-speed, low-light monster. I didn't think I'd ever say that about any 5D

 

[flanniganj]

Why the crappy post-processing? They all look very polished and no crispeness at all. Same with the 1d X samples. Who does post on these??

The camera. Canon tends to post unprocessed JPEG's strait out of the camera, and it looks like NR is cranked up as high as it will go on the high ISO shots. Nikon, on the other hand, seems to use manually processed JPEG's created from RAW, so they tend to look better than Canon samples. I think Canon is trying to be transparent about their camera's capabilities, which is something I'm appreciative of...but it makes it tough to really compare anything between brands when they do different things. The only way we are going to get a realistic comparison is to wait for someone to get their hands on all the competing bodies, take a shot of the same scene with each one, process each in the same way, and upload a visual comparison.

I disagree here. These were touched up beyond normal curves/exposure/sharpening. An example can be seen in the 2nd one. (02_cinc_big.jpg). Take a look along the ridge of trees in the distance. the tips of the trees near the sky are darker. It looks like they did a quick masking job to bring out the blue of the sky and didn't go into detail to remove out the tree tops. I'm not denying that it's minimal processing, but I agree with the original poster that it was done either hastily or poorly.

 

[V8Beast]

I ran one of Canon's ISO 25,600 samples through LR3 and indiscriminately cranked the noise reduction slider all the way to the right. IMHO, the result is a perfectly printable image and the resolution makes up for the loss of detail due to smearing in all but the largest of prints.

 

[Bonnau_Photo]

I have looked at all images from both sites posted here and see clearly that 12800 is useable. 25600 may work in a pinch with some added processing...

With that said I am happy to see Canon finally in the high ISO race and giving us something useable like this.. I know it will help my business a ton...
Now just think.. The 1Dx is supposed to be better

 

[Tuggem]

I hope the same people who has been analysing the new Nikons and made SNR measurements soon will get a chance to do the same with the new Canons.

 

[lubricus]

Why the crappy post-processing? They all look very polished and no crispeness at all. Same with the 1d X samples. Who does post on these??

The camera. Canon tends to post unprocessed JPEG's strait out of the camera, and it looks like NR is cranked up as high as it will go on the high ISO shots. Nikon, on the other hand, seems to use manually processed JPEG's created from RAW, so they tend to look better than Canon samples. I think Canon is trying to be transparent about their camera's capabilities, which is something I'm appreciative of...but it makes it tough to really compare anything between brands when they do different things. The only way we are going to get a realistic comparison is to wait for someone to get their hands on all the competing bodies, take a shot of the same scene with each one, process each in the same way, and upload a visual comparison.

I disagree here. These were touched up beyond normal curves/exposure/sharpening. An example can be seen in the 2nd one. (02_cinc_big.jpg). Take a look along the ridge of trees in the distance. the tips of the trees near the sky are darker. It looks like they did a quick masking job to bring out the blue of the sky and didn't go into detail to remove out the tree tops. I'm not denying that it's minimal processing, but I agree with the original poster that it was done either hastily or poorly.

If you go through the interface here:
http://cpn.canon-europe.com/content/education/technical/inside_canon_eos_5d_mark_iii.do
you'll find the image portfolio (third one down on the right).
Here, you can click on the exif data popup in the lower left.
If you do this, you'll see that the tree pic has an exposure setting of:
"HDR Art Vivid", so this is an example of the in-camera HDR functionality
There are a couple others shots using in-camera HDR:
06 (the night city bridge scene) is "HDR Art Standard"
14 (marina scene) is "HDR Natural"

15 (multiple headshots) gives "Multiple Exposure (Bright)" and
17 (jumping bike) gives "Multiple Exposure (Dark)"
18 (dancer) gives "Multiple Exposure"
... could these be in camera? ... that would be cool...

So, no, I don't think these images were necessarily retouched out of camera, although you were right in that the forest shot does not look normal!

 

[jwong]

Could it be that the black auras are due to the in camera chromatic aberration correction?