Approve the Cookies
This website uses cookies to improve your user experience. By using this site, you agree to our use of cookies and our Privacy Policy.
OK
Index  •   • New posts  •   • RTAT  •   • 'Best of'  •   • Gallery  •   • Gear  •   • Reviews
Guest
New posts  •   • RTAT  •   • 'Best of'  •   • Gallery  •   • Gear  •   • Reviews
Register to forums    Log in

 
FORUMS Canon Cameras, Lenses & Accessories Canon EOS Digital Cameras 
Thread started 03 Jul 2018 (Tuesday) 17:34
Search threadPrev/next
sponsored links
(this ad will go away when you log in as a registered member)

Canon AF zone focus algorithm issue

 
GregDunn
Goldmember
Avatar
1,287 posts
Likes: 127
Joined Mar 2013
Location: Indiana
     
Jul 03, 2018 17:34 |  #1

This is not so much a question as a comment, though if anyone has insight on the subject, feel free to weigh in.

I've progressively worked with a number of cameras over the last few years while doing sports photography: 7D, 7D mk II, 5D mk III, 1Dx, 1Dx mk II. I have noticed behavior which seems to contradict the manual and the Canon tech notes, and it doesn't seem to be my misreading of the descriptions.

Sometimes it's useful to set one of the various zone focus modes in AI Servo (I tend to use the smallest zone), when I'm shooting a small group of athletes. I know there are limitations, like keeping all the focus points within the area of the potential subjects, but occasionally it does what I want - focus on the group rather than my having to manually, relentlessly keep the center AF point on a subject which is in the middle of said group. However, Canon has published in multiple places that their zone AF tends to focus on the nearest subject in the focus point group, and after about 8-10 years of shooting with the various high-end cameras (literally tens of thousands of similar shots), I have come to the conclusion that this is not so.

If the contrast or pattern of a more distant object which happens to be within the zone is even a little higher than that of the closer object(s) - such as a patterned athletic uniform, a brightly colored banner or a set of bleachers - it will grab that object in preference to any closer one. It will do this even if it's barely visible between much closer subjects. It will often, perversely, focus on the rearmost in a group of athletes even when the zone boundaries are entirely within the desired group area (yes, I know that the AF points are larger than the ones shown in the viewfinder). If I had to quantify the behavior, I would say rather that it seems to focus randomly between the front and rearmost subjects, if there is no highly contrasty subject - and it always locks on the subject with highest contrast if it locks at all. (There's a separate issue with the cameras - all of them - indicating focus when there is nothing in focus, but that's not the subject of this observation) It doesn't seem to matter which of the several lenses I use either; from the 24-105 f/4 to the 70-200 f/2.8 II.

AF customization settings don't seem to matter (I've tried changing each of the 3 parameters separately as well as together); the best you can hope for is that there is nothing inside your zone which is behind your intended subject. Otherwise you're better off using single point AF and being careful (because if the AF point slips off the subject it too will jump to the background). Single + expansion points does the same thing even with tracking sensitivity set to minimum, because I've tried it - it seems to behave almost exactly the same as zone, though with fewer points to manage. It just loves higher contrast objects under the center AF point. It's not an AF microadjust issue either; single point, when I get it accurately on a subject, is dead on. What bugs me is that everyone just seems to repeat what Canon says about zones, but no one seems to actually have tested it and confirmed that it behaves the way they say. Every single one of the cameras I have owned which supports zone focusing has behaved the same, and no tweaking of the tracking parameters seems to help.


Canon 1Dx | 5D3 | 7D2 | 6D | 70-200L f/2.8IS | 70-200L f/4 | 24-70L f/2.8 | 24-105L f/4IS | 100-400L f/4.5-5.6IS | 17-55 f/2.8IS | 50 f/1.8 | 28-105 f/3.5-4.5 | 4x Godox AD360

  
  LOG IN TO REPLY
sponsored links
(this ad will go away when you log in as a registered member)
apersson850
Cream of the Crop
Avatar
11,953 posts
Gallery: 3 photos
Likes: 261
Joined Nov 2007
Location: Traryd, Sweden
     
Jul 05, 2018 03:32 |  #2

I don't too frequently use the zones available in my current two cameras (1DX and 1DX Mark II). I didn't use it too frequently in my previous 7D bodies either. Maybe for a reason.

But my feeling is that you are right. Good contrast, "low hanging fruit", tends to be favored by the camera. I would presume that if there are multiple targets with identical contrast (think laboratory conditions), then it would probably choose the nearest one. But such conditions tend never to exist in reality, which could explain the seemingly random subject choice.

My favorite sport is more of an individual one, so there single point with expansion(s) usually works well. Hence my limited experience with zones.


Anders

  
  LOG IN TO REPLY
Bassat
"I am still in my underwear."
8,075 posts
Likes: 2733
Joined Oct 2015
Post edited 8 months ago by Bassat.
     
Jul 05, 2018 05:43 |  #3

This completely explains the HUGE difficulties I'm having with my 80D while shooting incredibly tiny hummingbirds. Using single point or smallest zone, my hit rate is about 1%. The bg of my set up is much higher contrast than the tiny bird at 15 feet, but the bg is 200 yards away.

EDIT:
I have three hummingbird feeders around the yard. For one, the contrasting bg is 200' away, and is dark green corn. I get shots like the one posted, regularly; DOF is more of an issue than AF. The other has a large white house/barn/shop complex about 200 yards away. My AF hit rate there is just under 1%.


HOSTED PHOTO
please log in to view hosted photos in full size.




  
  LOG IN TO REPLY
TeamSpeed
01010100 01010011
Avatar
35,582 posts
Gallery: 102 photos
Best ofs: 2
Likes: 4994
Joined May 2002
Location: Cannelton
     
Jul 05, 2018 08:15 as a reply to  @ Bassat's post |  #4

It is no surprise really that high contract trumps distance to objects. This is one of the first points we bring up when folks wonder why their images are OOF on where they thought they put their AF point. 1) the AF point is usually larger than the little box you see as an indicator, coupled with 2) higher contrast areas will capture focus causing a potential OOF situation on the intended subject.

If you shoot a small subject material like an animal, etc, and there are bright objects on the periphery of said object, it can be frustrating watching the focus dive off into the distance on something that seems to have more defined detail than what you originally wanted.


Past Equipment | My Personal Gallery (external link) My Business Gallery (external link)

  
  LOG IN TO REPLY
mwsilver
Goldmember
3,821 posts
Gallery: 44 photos
Likes: 437
Joined Oct 2011
Location: Central New Jersey
Post edited 8 months ago by mwsilver.
     
Jul 05, 2018 14:15 |  #5

TeamSpeed wrote in post #18656727 (external link)
It is no surprise really that high contract trumps distance to objects. This is one of the first points we bring up when folks wonder why their images are OOF on where they thought they put their AF point. 1) the AF point is usually larger than the little box you see as an indicator, coupled with 2) higher contrast areas will capture focus causing a potential OOF situation on the intended subject.

If you shoot a small subject material like an animal, etc, and there are bright objects on the periphery of said object, it can be frustrating watching the focus dive off into the distance on something that seems to have more defined detail than what you originally wanted.

Isn't that a condition where use of Single-point Spot AF, which covers a smaller area than Single-point AF, comes into its own? While admittedly it doesn't work all the time, I've had good luck picking out small subjects between & behind other larger subjects. Granted the amount of contrast still has to be considered, but the size of the focus area around the point is decreased making it less likely you'll pick up other subjects which may be a bit closer. Of course the number of bodies that support Single-point Spot AF are limited. As an example, I'm not sure if the 80D has that feature. But when I'm in a tight focusing situation I've had significantly better luck using it than any other method.


Mark
Canon 7D2, 60D, T3i, T2i, Sigma 18-35 f/1.8, 30 f/1.4. Canon EF 70-200 L f/4 IS, EF 35 f/2 IS, EFs 10-18 STM, EFs 15-85, EFs 18-200, EF 50 f/1.8 STM, Tamron 18-270 PZD, B+W MRC CPL, Canon 320EX, Vanguard Alta Pro 254CT & SBH 250 head. RODE Stereo Videomic Pro, DXO PhotoLab, Elements 15

  
  LOG IN TO REPLY
TeamSpeed
01010100 01010011
Avatar
35,582 posts
Gallery: 102 photos
Best ofs: 2
Likes: 4994
Joined May 2002
Location: Cannelton
     
Jul 05, 2018 15:23 as a reply to  @ mwsilver's post |  #6

Yes spot AF shrinks sensitivity down somehow, and I use it quite a bit. However you sacrifice AF speed when using that, so there are always compromises.


Past Equipment | My Personal Gallery (external link) My Business Gallery (external link)

  
  LOG IN TO REPLY
russbecker
Member
232 posts
Gallery: 27 photos
Likes: 336
Joined Jun 2007
Location: Central PA, USA
Post edited 8 months ago by russbecker.
     
Jul 05, 2018 20:00 |  #7

I noticed this same issue more than a decade ago using my 40D followed by my 7D for shooting indoor sports such as gymnastics. The most maddening was having an athlete up on the uneven bars and having the camera (doesn't matter which one) maddening lock in on the cinder block wall behind her, even though the block pattern is of apparently low contrast -- it still has a pattern which the AF system is attracted to like flies to honey.

Never used the 7D2 for indoor sports, but quite a bit for T/F and noted the same issue with regular pattern backgrounds like chain link fences. As TEAM SPEED pointed out, you can mitigate the effect somewhat by going to SPOT AF which will cost you some speed and sensitivity. 80D doesn't have this mode so no help there.

You just have to be aware of the phenomena and try to stay away from those situations.


7D2 | 7D | 80D | 40D | 100-400 f/4-5.6 IIL | 300 f/4 L | 70-200 f/2.8 IIL | 70-200 f/4L | 135 f/2 L | 85 f/1.8 | 100 f/2 | 60 f/2.8 macro | nifty-fifty | 10-22 f/3.5-4.5 | Tamron 150-600 | Tamron 17-50 f/2.8 | Sigma 30 f/1.4 | Sigma 50-150 f/2.8 | Sigma 120-400

  
  LOG IN TO REPLY
Pekka
El General Moderator
Avatar
17,711 posts
Gallery: 35 photos
Best ofs: 7
Likes: 1480
Joined Mar 2001
Location: Hellsinki, Finland
     
Jul 06, 2018 04:34 |  #8

GregDunn wrote in post #18655828 (external link)
I would say rather that it seems to focus randomly between the front and rearmost subjects, if there is no highly contrasty subject - and it always locks on the subject with highest contrast if it locks at all.

This has been so from the start of Canon DSLR lineup (since D30, ca. 2001). The AF system works solely on contrast and it does not understand depth at all. The only hardware option to get depth info would be to make the AF scan (move USM) the whole range and check for focusing distances from the lens motor, but that info is very rough estimate. Until we get depth-aware cameras, like iPhone X, and we can just tell it to "keep focus on closest person", the photographer needs to assist the AF pretty much constantly and focus lock-recompose a lot when there are strong contrasts in front or behind the intended subject.

The "locked to nothing" must be related to fact that AF engine must make a decision very fast in order to satisfy current speed demands, so if it does not get it 100% it accepts 80% to avoid back-forth pumping.


The Forum Boss, El General Moderator
AMASS 2.1 Changelog (installed here now)

  
  LOG IN TO REPLY
BigAl007
Cream of the Crop
Avatar
7,962 posts
Gallery: 542 photos
Best ofs: 1
Likes: 1597
Joined Dec 2010
Location: Repps cum Bastwick, Gt Yarmouth, Norfolk, UK.
     
Jul 06, 2018 06:38 |  #9

Pekka wrote in post #18657228 (external link)
This has been so from the start of Canon DSLR lineup (since D30, ca. 2001). The AF system works solely on contrast and it does not understand depth at all. The only hardware option to get depth info would be to make the AF scan (move USM) the whole range and check for focusing distances from the lens motor, but that info is very rough estimate. Until we get depth-aware cameras, like iPhone X, and we can just tell it to "keep focus on closest person", the photographer needs to assist the AF pretty much constantly and focus lock-recompose a lot when there are strong contrasts in front or behind the intended subject.

The "locked to nothing" must be related to fact that AF engine must make a decision very fast in order to satisfy current speed demands, so if it does not get it 100% it accepts 80% to avoid back-forth pumping.


That would be true for cameras when in liveview, at least until the advent of the Dual Pixel sensor, but not for the normal autofocus via the OVF. For that the camera does need an area of contrast to focus, but the focus movement needed is derived from phase differences in the light coming from different sides of the lens. The great thing about the phase difference method is the fact that this signal is able to provide information on both how far from focused the lens is, and in what direction. It's this ablity to detect the distance and direction required to achieve focus that makes PDAF so fast to operate. In some cameras it was suspected that the AF system simply did a single look/move cycle then fired the shutter, rather than doing look/move/look to confirm the focus is correct.

The aperture limit for lenses for PDAF is also related to this, since in general the wider the aperture the longer the baseline you are working with to generate the phase shift. This is why the most accurate AF sensor locations are in the center of the frame. The AF sensors will be the same, but the central location allows for the biggest phase shift at any aperture.

If there were say two different areas of contrast present at a focus point, at different distances from the focal plane, then the camera will see two different phase shifts. The camera should then be able to select the phase shift that represents the closest of the potential possible focus distances. This is the advertised action for the AF system, of course the programming could be set up to favour any distance options you like.

The problem with overlapping signals is that as far as the AF sensor is concerned the level of the signal coming from a further point may well be significantly higher than for the closer points. In this situation it may well be that the big signal swamps the smaller one, and the camera is then unable to "see" it as a valid signal. So you get fous in the "best" target, not the closest one as you should.

Personally I think that the AF we have now is incredible, what with it working down to -3EV, it is actually better than the metering system at that point. Still give it a complicated multi subject/distance scene to focus on and it's not surprising that it occasionally struggles. It strikes me that it is much better than being stuck with manual focus only, although we managed even with that.

Alan


My Flickr (external link)
My new Aviation images blog site (external link)

  
  LOG IN TO REPLY
Bassat
"I am still in my underwear."
8,075 posts
Likes: 2733
Joined Oct 2015
     
Jul 06, 2018 11:05 as a reply to  @ BigAl007's post |  #10

Physics on the net! ߘ




  
  LOG IN TO REPLY
digital ­ paradise
How do I change this?
Avatar
14,327 posts
Likes: 7652
Joined Oct 2009
     
Jul 06, 2018 11:08 |  #11

I shot for 5 years using centre point only with my 7D. Since the 7D2 and 5D4 I use both Zone and Auto frequently but I don't use Large zone. I find Zone is pretty good and I don't see it being mentioned very often. I tend not to use it with very busy environments because yes it can loose subject focus. I have yet to try iTR with my 5D4 which is more advanced than the 7D2.

I did some tests with the 5D4 in auto. A deer, in the fall with brown bush behind it. It was pretty good at focusing on the closest object and it tracked it well. The deer was at about a 45 degree angle and had it's butt facing towards me and it focused on the butt. It did this repeatedly with any part of the deer that was closest. The bush was behind it so probably a different story with it in the bush. Zone is useful but it does have limitations.

I have be told that if you shoot a bird on water Auto will focus on the water in front of the bird. I have yet to find this. It seems centre weighted to me and some days I think it is sensitive to movement. It would make sense as typically we focus most things in the centre of the frame. It is interesting that when I use Zone or Auto for BIF it pretty much does not focus on the wing closest to me but on the body. I think Zone is better for BIF because I can't keep the AF point on the eye so what is the difference what part of the bird AF is acquiring and depending on the distance DOF is not an issue.

It works pretty good when the initial AF point is activated. After it acquires AF it drops out and becomes part of the array and appears again when you take your finger off the BBF or shutter.

I have seen some sports videos with the 5D4 using iTR and facial recognition but I have not tried it yet.

As for the Centre point dropping focus I use Case 2 with the TS set to -2. If the centre point goes off target you get about a second tp get back on the original target before it starts to focus in something else. May not seem like much but at 10 to 16 FPS it can make a world of difference. Case 3 is the opposite of Case 2 where it refocuses quickly into the other subject. Expansion solves that but you need to keep one of the expansion points on the subject.

I did some controlled tests to see how expansion worked and it does as advertised. I also did a controlled test to see if changing Case numbers effected AF acquisition speed. It didn't. It all about contrast and we have no fine tuning control over that. This ties back to your statement about Zone picking up something with better contrast. I think iTR might be a helpful solution to that. I'm going to have to give it a good try out this summer.

Since getting my 7D2 I have only used the zone AF modes in Cases 5 and 6 only. Of course you can apply those parameter settings to many Case number. I have never felt that an increase in AF switching did very much for me in expansion mode. I did some searches and found an Art Morris (I know he went to Nikon) blog and made for Canon video where he said it didn't do anything. Don't know if he is right or wrong and I don't really care but I did agreed it didn't to much, for me anyway.

Then the 1DX II 2 came and the AF guide says this and I agree with it. The 1DX II guide also has the best explanation for Case 4 which confirmed Accel/Decel is for front and back movement.


HOSTED PHOTO
please log in to view hosted photos in full size.


Image Editing OK

Website (external link) ~ Buy/Sell Feedback

  
  LOG IN TO REPLY
apersson850
Cream of the Crop
Avatar
11,953 posts
Gallery: 3 photos
Likes: 261
Joined Nov 2007
Location: Traryd, Sweden
     
Jul 09, 2018 06:32 |  #12

The documentation for the 1DX Mark II also confirms that AF switching, as normally applied by Cases 5 & 6, also applies to single AF point with expansion points. But as there's less leeway for choosing points in these modes, it's logical that the best use of these cases is where the camera has the most AF points to play with.


Anders

  
  LOG IN TO REPLY
digital ­ paradise
How do I change this?
Avatar
14,327 posts
Likes: 7652
Joined Oct 2009
     
Jul 09, 2018 09:09 |  #13

I did read that as well and since I follow Canon documentation pretty closely I'm good with it. We talked about this before and I take Art's comments with a grain of salt. He probably just meant it did not have a significant impact. That was what I was really looking for when I did my searches on that subject. I probably did not explain myself well a few years ago but I decided back then how I would utilize those 2 cases and have been shooting that way since. Canon keeps updating documentation and the 1DX II AF guide was the most informative for me.


Image Editing OK

Website (external link) ~ Buy/Sell Feedback

  
  LOG IN TO REPLY
digital ­ paradise
How do I change this?
Avatar
14,327 posts
Likes: 7652
Joined Oct 2009
Post edited 8 months ago by digital paradise. (3 edits in all)
     
Jul 09, 2018 09:36 |  #14

Now with my 400 DO II and 2X I'm going to need a 1DX II :-( Going to have to sell another kid :-) I can BIF but it is not easy. Then again I aways like a challenge.

AF speed

A sharp lens that is slow to focus is of no use to me as I like to capture birds in flight. Prior to the introduction of the EOS-1D X, Canon electronically slowed down the AF servo drive speed when either 1.4X or 2X extenders were attached to account for the loss of light, that is as 50% with the 1.4X extender and 75% with the 2X extender. The significant reduction in servo drive speed, especially with the extender 2X, limited its application for bird in flight photography. Canon changed this equation with the introduction of the EOS-1D X and series II super-telephoto lenses. When using a series II super-telephoto lens and a series III extender, the EOS-1D X user will not notice a reduction in the servo drive speed with either 1.4X or the 2X extender. (Note that is not true for any other camera or lens combination that I have used). This means that when attached to an EOS-1D X, the EF 400mm f/4 DO IS II plus EF Extender 2X III can potentially be a great combination for shooting birds in flight.


http://arihazeghiphoto​graphy.com …ttle-of-the-light-lenses/ (external link)


Image Editing OK

Website (external link) ~ Buy/Sell Feedback

  
  LOG IN TO REPLY
GregDunn
THREAD ­ STARTER
Goldmember
Avatar
1,287 posts
Likes: 127
Joined Mar 2013
Location: Indiana
     
Jul 10, 2018 00:53 |  #15

Glad to see that others are seeing pretty much the same thing as I am (though sad that it's apparently an accurate description of the zone algorithm).

I agree - with PDAF giving the camera distance info and the high powered processors now installed in the higher-end Canons, they should be able to bias focus to the front - but it doesn't seem to be working like that in zone mode unless you have a nice bland background. I shot an event last weekend, and tried to find a combination that worked best for what I was doing. Single-point (I don't use spot because I'm always in Servo with moving subjects and another thing Canon says is not to do that) worked the best when I was able to track the subject. I have tried cases 5 and 6 but the problem is that the athletes move generally toward me and do not accelerate or decelerate, other than side to side in some cases where I need to reacquire. Increasing the accel/decel parameter frequently causes my cameras to overshoot focus when I switch subjects (I am "pumping" the AF button to make it reacquire, rather than letting the camera decide when I am switching subjects). After lots of frustration, that parameter has been set as far to the left as I can get away with.

I had success with zone a few times when there was a wall of athletes blocking the background (and I had iTR turned on), but it failed plenty of times; single point expanded was about the same, almost never as good as single point. Once again, the expansion points love to find high contrast in a tiny corner of the measurement area and grab onto it. Proof that the AF point is larger than the square in the viewfinder was shown to me many times, when the square was entirely over the subject but focus was achieved on something just a little further beyond.

I am starting to believe that if the AF point can't find quite enough contrast, the camera will eventually give up and say it's close enough to focus that it allows the shutter to release. I do have focus priority set on all the AF options, but all of my bodies have been known to shoot when there is literally nothing in focus anywhere near the AF point(s). I will release the shutter and sometimes the camera will jump out of focus just as the mirror starts to come up, even though it was apparently in focus just before. I've tried with and without IS enabled to see if that's a factor, but it doesn't seem to be - if anything, "on" helps me hold the AF point more accurately so the result is a net improvement. Fortunately, these incidents are pretty rare.

I can't really complain too much, as bodies like the 1Dx2 are far better overall than my old 7D or even 5D3 - but it's still frustrating because I can't hold the AF point on my target 100% of the time, or guarantee that a jersey with low contrast designs will hold the camera's interest long enough. The only trick I can share is that I move the center point up a couple of clicks so that the athletes' faces are closer to it, and that seems to help. But of course there are gaps between their heads which give the camera opportunities to find background contrast too. It's a game of try, observe, change a setting, try again - and different venues, uniforms, and lighting all change the rules each time. With single point, I am at least confident that all the mistakes are mine, so that's probably why it's the least frustrating mode to use. :-)


Canon 1Dx | 5D3 | 7D2 | 6D | 70-200L f/2.8IS | 70-200L f/4 | 24-70L f/2.8 | 24-105L f/4IS | 100-400L f/4.5-5.6IS | 17-55 f/2.8IS | 50 f/1.8 | 28-105 f/3.5-4.5 | 4x Godox AD360

  
  LOG IN TO REPLY
sponsored links
(this ad will go away when you log in as a registered member)

1,907 views & 8 likes for this thread
Canon AF zone focus algorithm issue
FORUMS Canon Cameras, Lenses & Accessories Canon EOS Digital Cameras 
AAA
x 1600
y 1600

Jump to forum...   •  Rules   •  Index   •  New posts   •  RTAT   •  'Best of'   •  Gallery   •  Gear   •  Reviews   •  Member list   •  Polls   •  Image rules   •  Search   •  Password reset

Not a member yet?
Register to forums
Registered members may log in to forums and access all the features: full search, image upload, follow forums, own gear list and ratings, likes, more forums, private messaging, thread follow, notifications, own gallery, all settings, view hosted photos, own reviews, see more and do more... and all is free. Don't be a stranger - register now and start posting!


COOKIES DISCLAIMER: This website uses cookies to improve your user experience. By using this site, you agree to our use of cookies and to our privacy policy.
Privacy policy and cookie usage info.


POWERED BY AMASS forum software 2.1forum software
version 2.1 /
code and design
by Pekka Saarinen ©
for photography-on-the.net

Latest registered member is notinspired
1091 guests, 360 members online
Simultaneous users record so far is 15144, that happened on Nov 22, 2018

Photography-on-the.net Digital Photography Forums is the website for photographers and all who love great photos, camera and post processing techniques, gear talk, discussion and sharing. Professionals, hobbyists, newbies and those who don't even own a camera -- all are welcome regardless of skill, favourite brand, gear, gender or age. Registering and usage is free.