5D3suffer blackdots(EDIT:my mistake,just poorly tuned sharpening algo4incamJPGS

[LetTheRightLensIn]

Re: Why was the 5D3 black outlines thread locked???

With the 5D Mark II, the black dots were clearly visible on a 100% crop. In the sample image in the 5D Mark III black dot thread, the dots are barely visible at 400%.

wow, that's a valid point, i didn't know about the 5D2.. was it a hardware or software issue?

The problem on the 5D Mark II was corrected with a firmware update.

Umm no these are visible at 100% too (I mean how did you think people noticed it again?) and are actually easier to spot and seem to occur under less extreme conditions.

 

[DJL329]

Re: Why was the 5D3 black outlines thread locked???

With the 5D Mark II, the black dots were clearly visible on a 100% crop. In the sample image in the 5D Mark III black dot thread, the dots are barely visible at 400%.

wow, that's a valid point, i didn't know about the 5D2.. was it a hardware or software issue?

The problem on the 5D Mark II was corrected with a firmware update.

Umm no these are visible at 100% too (I mean how did you think people noticed it again?) and are actually easier to spot and seem to occur under less extreme conditions.

Because they're looking for problems where none exist, that's why.

I found an example of the black dot issue from a 100% crop using the 5D Mark II:

http://www.kenrockwell.com/canon/slrs/5d-mark-ii/black-pixels.htm

And here's the 400% crop example from the 5D Mark III posted in the other thread:

So, as anyone can clearly see, there is no comparison. And that's why the other thread was closed.

 

[LetTheRightLensIn]

Re: Why was the 5D3 black outlines thread locked???

With the 5D Mark II, the black dots were clearly visible on a 100% crop. In the sample image in the 5D Mark III black dot thread, the dots are barely visible at 400%.

wow, that's a valid point, i didn't know about the 5D2.. was it a hardware or software issue?

The problem on the 5D Mark II was corrected with a firmware update.

Umm no these are visible at 100% too (I mean how did you think people noticed it again?) and are actually easier to spot and seem to occur under less extreme conditions.

Because they're looking for problems where none exist, that's why.

I found an example of the black dot issue from a 100% crop using the 5D Mark II:


And here's the 400% crop example from the 5D Mark III posted in the other thread:


So, as anyone can clearly see, there is no comparison. And that's why the other thread was closed.

It's not clear for sure, but it shows some similarities, it seems to appear only along the brightest of all edges. What about the chunk of snow on the snowboarder pic where the left side is all lined in black?

Since when do stars have black circles around them on a blue background?
When do chunks of ice get outlines in black?
Something weird is going on.

And no people were not going around at 400% view looking for problems. The second I opened the image (normally) I was boom what is that??? And the same happened when others saw it.

Now maybe it's just that the new sharpening algorithm had white halos undertuned and black halo over-tuned and it won't be there in the RAW file, but at the least it is a badly tuned sharpening algorithm in the in cam jpg engine. At worst it is some form the black dot issue again.

Actually looking more closely some of the outlines are at full 0,0,0 so maybe it's not quite the same or maybe they only partially fixed or maybe it's just something screwy in their sharpness+NR algorithm for in cam jpgs (which won't affect RAW files one bit, unlike the true black dots issue). It's still a bit bizarre to use sample images where the algorithms have been so poorly tuned to this point even if it is just the latter. And it may help to explain why some are saying some of the samples just have such a weird, weird look even outside of the often heavy handed NR.

It does show up pretty much only near extreme whites though, but yeah since some of the outlines are not 0,0,0 it might have to be something else (or a partial fix?).

 

[pedro]

Re: 5D3 black outlines - something weird with in cam jpg sharpening+NR (or maybe

These samples were taken with a pre-production unit, right? I guess Canon are aware of it and might have it fixed once the first batch hits the street, or yes a firmware update will do.

 

[thenickdude]

Re: 5D3 black outlines - something weird with in cam jpg sharpening+NR (or maybe

That snow example has (to my eye) a white halo of similar magnitude on the inside of the edge, so it really looks like a sharpening issue instead of some sensor problem.

 

[LetTheRightLensIn]

Re: 5D3 black outlines - something weird with in cam jpg sharpening+NR (or maybe

That snow example has (to my eye) a white halo of similar magnitude on the inside of the edge, so it really looks like a sharpening issue instead of some sensor problem.

Yeah I am starting to think that it may just be that they tuned the sharpening weirdly.

I was once working with a sharpning algorithm where you could tune the white and the black halo oveshoot and it seems they just had it tuned very weirdly at that point (hopefully not still, although I do mostly shoot RAWs so it's not end of the world, but it would be a bit nasty for the times you do want to shot in cam jpgs) to where they had strong outer black haloes.

 

[altenae]

Re: 5D3 black outlines - something weird with in cam jpg sharpening+NR (or maybe

Check Digvis sharpening example here:

http://forums.dpreview.com/forums/readflat.asp?forum=1032&thread=40798618

 

[K-amps]

Re: 5D3 black outlines - something weird with in cam jpg sharpening+NR (or maybe

Check Digvis sharpening example here:

http://forums.dpreview.com/forums/readflat.asp?forum=1032&thread=40798618

Looks like a sharpening software issue..

 

[eeek]

Re: 5D3 black outlines - something weird with in cam jpg sharpening+NR (or maybe

By the way- that Ken Rockwell article is wonderful. It really should be read before this thread is.

 

[LetTheRightLensIn]

Re: 5D3 black outlines - something weird with in cam jpg sharpening+NR (or maybe

Check Digvis sharpening example here:

http://forums.dpreview.com/forums/readflat.asp?forum=1032&thread=40798618

Looks like a sharpening software issue..

Yes, i'm 99% sure I was wrong. At first glance it looked like the black dots issue since it was seen only around the blown out white parts and most sharpening plus NR algorithms are not to rob so many details and still leave such major halos and they are more often lighter than darker but yeah it seems they just over-tuned the black halo settings.

They are using a nastily tuned sharpening algorithm though that isn't tuned well and is much too prone to suffering from black halo overshoot, they appear to have tuned the white halos away nicely but over-tuned the black halos. So lots of the in cam jpg samples have a slightly weird look to them.

For RAW shooters this won't matter in the slightest.

For JPG shooters it will matter unless they have fixed it since these images were shot a few months ago.

It is a bit sloppy that Canon is using such terrible quality samples. They always do, but these are, in many ways the worst sample set yet for any of their bodies. Very bizarre.

 

[LetTheRightLensIn]

Re: 5D3 suffer black dots (EDIT: just poorly tuned sharpening+NR in cam jpg algo)

Anyway, sorry for the false alarm.

(they should fine tune their sharpening algorithm for in cam jogs though, assuming they have not in the months since the samples were taken, they are overshooting the black halos too aggressively)

 

[Fleetie]

Re: 5D3 suffers from the black dots problem of the early 5D2

As an experienced astro-photographer, I can tell you that black halos around stars are quite common with almost any type of camera, including dedicated astro-cameras which are basically chip only. Stars, especially near the edge of a frame, are possibly the single most demanding subject in any kind of photography - period. They are small, should be perfectly round, should not exhibit color fringes or halos. They are very demanding of optics and chips and processing. Any small flaw in any of the above can and will cause problems, sometimes these issues show up as dark halos.

Yes, they are essentially examples of the Dirac Delta Function, a spike with infinite height (brightness, here), zero width (their width cannot be resolves because they are so small because so distant (except a very few, like Betelgeuse)), and unit area.

http://en.wikipedia.org/wiki/Dirac_delta_function

Applying a Delta function as input to any system (amplifier) and measuring the output from that system will in theory tell you everything about the response of that system.

In any non-ideal system, you expect to see "crap" coming out in response; that "crap" contains information about the response of the system.

 

[LetTheRightLensIn]

Re: 5D3 suffers from the black dots problem of the early 5D2

As an experienced astro-photographer, I can tell you that black halos around stars are quite common with almost any type of camera, including dedicated astro-cameras which are basically chip only. Stars, especially near the edge of a frame, are possibly the single most demanding subject in any kind of photography - period. They are small, should be perfectly round, should not exhibit color fringes or halos. They are very demanding of optics and chips and processing. Any small flaw in any of the above can and will cause problems, sometimes these issues show up as dark halos.

It's more just the sharpening since it also happens to same degree for the large chunks of flying snow, etc.

But again, sorry, my mistake, it does not appear to be the 5D2 black dots thing again. That left 0,0,0 signal in RAW files and behaved a bit differently.

Sorry again.

(although my post and others who posted on it elsewhere do bring up the valid point that the in cam jpg sharpening+ NR algos are not, or at least were not back when these were taken, tuned well, it doesn't hurt to bring that to attention)

 

[SiliconVoid]

It does not really matter if it is showing up.
The two things to keep in mind is one, it is unfortunately part of digital photography (as opposed to film) and all cameras do it. The manufacturers do all kinds of digital artifact mapping for each sensor design. If the compression algorithm is not addressing it then Canon will fix it. Which brings us to the second, Canon tweaked the algorithm in the 5DmkII and will fix it in the mkIII - if it even actually exists beyond those early preproduction test units.

I know everyone is excited, and need something to do until the mkIII begins shipping and/or until they save up their money but this is not productive in any way. Canon has a long list of things they know to look for based on the lineage of the 5D series.
- I suggest you go outside and take some pictures.