The first noise test results from the Canon EOS R3 are in

AlanF

Desperately seeking birds
CR Pro
Aug 16, 2012
12,343
22,518
I always feel like my TSE 24 II has more DR than any other lense ever used on my 5div.
Do also other experienced this phenomenon?

I mean when I changed lense in a landscape situation the light range that this lense can cover is little higher than with other lenses.
Basically the histogram graph is with the same body and same scene a little bit more compressed on the TSE 24 II.
I believe, this is the main reason why photos taken with the TSE 24 II make me smile each time I see them after the shot
Dynamic range is actually calculated for images enlarged to the same size and distance of viewing, roughly equivalent, from memory, to about an 8"x10" print at arms length. That's why an APS-C has lower DR because it has to be enlarged more and medium format higher DR than FF because it has to be enlarged less.
 
Upvote 0
Colours seem to shift a lot more magenta in the R5 and R6 compared to the R3 - maybe that's just an incorrect WB on the R3 though.

R3 has improved WB over R5/6. This may be the first time AI/deep-learning has been used to improve WB.

Quote:
Canon says the R3's Auto White Balance has been enhanced by machine learning, to help it better interpret scenes, and not get thrown off by, for example, landscape images dominated by greenery.

 
  • Like
Reactions: 1 user
Upvote 0
Jul 28, 2015
3,368
570
R3 has improved WB over R5/6. This may be the first time AI/deep-learning has been used to improve WB.

There is no 'AI' or 'Deep learning'. It is a sales buzzword for ever more sophisticated programming (like 'nano' was a few years ago).
To 'learn' or 'apply AI' the camera program needs feedback to know what it is doing right and what it is doing wrong. No such feedback is given to any camera that I know of.
 
Upvote 0

Deleted

7D2
Sep 30, 2021
111
340
There is no 'AI' or 'Deep learning'. It is a sales buzzword for ever more sophisticated programming (like 'nano' was a few years ago).
To 'learn' or 'apply AI' the camera program needs feedback to know what it is doing right and what it is doing wrong. No such feedback is given to any camera that I know of.
I can only guess that it refers to how the algorithm was built in the first place. By utilising many image on a dedicated platform the algorithm that wil ultimately be used in camera can be built and tuned. So while it may or may not be happening in camera, it was almost certainly artificial learning that created the parameters in the first place.
 
  • Like
Reactions: 1 user
Upvote 0
Jul 21, 2010
31,088
12,851
I only shoot raw. Pardon my ignorance, is there a camera setting that reduces noise in raw files?
Only with long exposures (dark frame subtraction).

My point is that if you open a RAW file in most converters (even Preview on a Mac), there is NR applied by default. For the TDP noise tests, NR is set to off prior to conversion, so they look noisier than what people typically see when opening their RAW files.
 
  • Like
Reactions: 1 users
Upvote 0

Deleted

7D2
Sep 30, 2021
111
340
Only with long exposures (dark frame subtraction).

My point is that if you open a RAW file in most converters (even Preview on a Mac), there is NR applied by default. For the TDP noise tests, NR is set to off prior to conversion, so they look noisier than what people typically see when opening their RAW files.
It was my understanding that in some raw converters the NR applied in the preview was based on the profile set in camera. Is that still the case with the R3?
 
Upvote 0
There is no 'AI' or 'Deep learning'. It is a sales buzzword for ever more sophisticated programming (like 'nano' was a few years ago).
To 'learn' or 'apply AI' the camera program needs feedback to know what it is doing right and what it is doing wrong. No such feedback is given to any camera that I know of.
I think we all know there is not an AI literally living in the camera :) The AI programming is done in the lab, and it's used to create the algorithm used in the firmware.
 
Upvote 0
Only with long exposures (dark frame subtraction).

My point is that if you open a RAW file in most converters (even Preview on a Mac), there is NR applied by default. For the TDP noise tests, NR is set to off prior to conversion, so they look noisier than what people typically see when opening their RAW files.
Thanks. That makes sense.
 
Upvote 0

koenkooi

CR Pro
Feb 25, 2015
3,569
4,109
The Netherlands
I think we all know there is not an AI literally living in the camera :) The AI programming is done in the lab, and it's used to create the algorithm used in the firmware.
I'm looking forward to a Canon camera that is able to learn from the manual adjustments made by me, so it will know that for a bee shaped object I'd like to focus on the jaws or eyes.

Once you have enough processing power on the device, you can do a lot of interesting things. Apple has published a bunch of papers on how to augment facial recognition with clues from the upper body, like clothing. It would be nice to have something like that so you can track a specific person moving through a crowd. Or cheat at where's Waldo :)

I'm also surprised that Canon doesn't seem to use the depth information they get from DPAF, I bet that would help a lot with not focussing on the foliage behind the bird.
 
  • Like
Reactions: 1 user
Upvote 0

FramerMCB

Canon 40D & 7D
CR Pro
Sep 9, 2014
481
147
56
I always feel like my TSE 24 II has more DR than any other lense ever used on my 5div.
Do also other experienced this phenomenon?

I mean when I changed lense in a landscape situation the light range that this lense can cover is little higher than with other lenses.
Basically the histogram graph is with the same body and same scene a little bit more compressed on the TSE 24 II.
I believe, this is the main reason why photos taken with the TSE 24 II make me smile each time I see them after the shot
Sounds like the TSE 24mm Mk II has better light transmission?
 
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
I always feel like my TSE 24 II has more DR than any other lense ever used on my 5div.

Basically the histogram graph is with the same body and same scene a little bit more compressed
Dynamic range of a scene may be affected by the lens. For example, anything that lowers contrast, such as flare or ghosting, will compress the histogram. Similarly, video is shot with very flat profiles that are super low contrast to compress the DR of a scene into a tighter histogram. But that is done digitally after the image is created on the sensor.

Maybe your lens isn't as punchy as your others.

But that's not what we're concerned with when talking about a cameras DR for stills. The sensor is a bottleneck independent of the lens. Any kind of contrast reduction in a lens that I'm aware of comes from light bleeding into areas where it should not be. So it is degrading the image, even if the actual contrast can easily be compensated for in post.

So even though your shadow areas may get a bit brighter with enough flare/ghosting in your image, it doesn't actually mean you're getting any additional detail in those shadows. And therefore not any real gain in dynamic range.
 
Upvote 0