Page 2 of 3 FirstFirst 123 LastLast
Results 16 to 30 of 32

Thread: Keying issue with Mini SDI AE2 that I never had with TC40

  1. #16
    Registered User stanarthur's Avatar
    Join Date
    Jun 2013
    Location
    Florida
    Posts
    50
    the overly-lit background observation is spot on, I think. It calls for more experimentation. One of the reasons it's so bright is because it's ProCyc, the brightest green screen vendor in the biz. I can definately control that and focus more light on the talent. That's what's next.
    TriCaster Mini SDI Advanced Edition 2 Build 5-1-180323C
    Talkshow

  2. #17
    Registered User stanarthur's Avatar
    Join Date
    Jun 2013
    Location
    Florida
    Posts
    50
    The good news is that I was able to get rid of the white line. The bad news is what it took to do it and trying to figure out how I'm going to accomplish it for all three cameras during a live show. The solution was a mixture of most of your advice regarding image quality (the noise issue) and lighting. I switched out the lens for a faster one, a T1.5 cine, reduced the camera gain from +18dB to +6dB and the noise disappeared. Additionally, I moved my four soft boxes closer to the talent and turned off the fluorescent fixtures that were lighting the green screen. When keying, I turned the softness to zero and turned the tolerance up until the white line was gone (about 48%), then brought the softness up to about 12%. This was by far the best result.
    TriCaster Mini SDI Advanced Edition 2 Build 5-1-180323C
    Talkshow

  3. #18
    LiveSet Making Machine joseburgos's Avatar
    Join Date
    Feb 2003
    Location
    NYC
    Posts
    3,307
    Have you tried using the Effect Show Alpha to take a look at your key? You can load it into one of your M/E's and input A is only used by the Effect. I usually teach this to clients who use Virtual Sets when I train them to help them work on getting a great key
    Jose Burgos
    NewTek Certified Trainer
    NewTek Certified on all TriCaster's
    NewTek Training & Certification Testing
    www.burgosfx.com
    FaceBook
    Twitter @NYTriCaster

  4. #19
    Registered User stanarthur's Avatar
    Join Date
    Jun 2013
    Location
    Florida
    Posts
    50
    I'll give it a try, Jose if I can figure out how to load an effect into an ME.
    TriCaster Mini SDI Advanced Edition 2 Build 5-1-180323C
    Talkshow

  5. #20
    Registered User
    Join Date
    Apr 2008
    Location
    Racine
    Posts
    1,299
    The "Show Alpha" and other goodies are found where the virtual sets are, in same list

    Thanks

    Jeff
    Jeff Pulera
    Streaming Broadcast Solutions - Newtek Elite

    TriCasters: Mini with AE, TC1
    Camera: Sony PMW-X70 4K
    Controllers: All variety of XKeys
    PTZ: Newtek NDIHX-PTZ1

  6. #21
    LiveSet Making Machine joseburgos's Avatar
    Join Date
    Feb 2003
    Location
    NYC
    Posts
    3,307
    Jose Burgos
    NewTek Certified Trainer
    NewTek Certified on all TriCaster's
    NewTek Training & Certification Testing
    www.burgosfx.com
    FaceBook
    Twitter @NYTriCaster

  7. #22
    Registered User
    Join Date
    Sep 2017
    Location
    Sweden
    Posts
    6
    Quote Originally Posted by stanarthur View Post
    The good news is that I was able to get rid of the white line. The bad news is what it took to do it and trying to figure out how I'm going to accomplish it for all three cameras during a live show. The solution was a mixture of most of your advice regarding image quality (the noise issue) and lighting. I switched out the lens for a faster one, a T1.5 cine, reduced the camera gain from +18dB to +6dB and the noise disappeared. Additionally, I moved my four soft boxes closer to the talent and turned off the fluorescent fixtures that were lighting the green screen. When keying, I turned the softness to zero and turned the tolerance up until the white line was gone (about 48%), then brought the softness up to about 12%. This was by far the best result.
    We’ve had the same issues with bright lines on just the right side. Part of the problem is definitely ProCyc, since the green is so bright it’s very hard to manage spill if you don’t have a huge studio with lots of distance between the person and the walls. We ended up repainting our studio in regular Chroma Key green, which is about two full stops darker. That made spill much easier to handle, made setting the proper exposure for the foreground and green screen easier and allowed us to actually use the screen lights to get rid of some shadows and a generally more even result. It did come with a negative side effect of some extra noise in the key.

    What it didn’t do was to get rid of the line on the right side, only now it’s a dark line that shows on bright clothes over light backgrounds. Investigating further, I discovered that some of the problem with ugly outlines can be explained by chroma subsampling (422) which makes high contrast edges “bleed” due to the lower resolution of the color channels. This can easily be observed outside chroma keying as well if you try to put red text over a green background and export it to any codec that uses 422 (most of them). Other causes for this edge bleed is poor quality internal downscaling from 4K to 1080p in many cameras and chromatic aberration, especially with cheaper lenses or when using speed booster type lense adapters.

    That still didn’t explain why the lines (dark or bright) only showed up on the right side. Applying a simple chroma key in any editing software resulted in a small but evenly distributed and therefore less noticeable outline. Using the Show Alpha M/E preset I captured the alpha matte generated by our Tricaster and compared that to a carefully matched key done in After Effects. This showed that the alpha matte generated by the Tricaster is actually shifted slightly to the right! This of course means that no matter what you do, more of the background will always be visible to the right of the keyed subject, leading to the infamous right side outline.

    I’ve been trying for quite some time to get NewTek to take this seriously, filing bug reports and providing material for testing. I’ve seen other threads here reporting this issue, but the posters eventually get shot down for not using the proper lighting technique, having the wrong camera settings or something else reflecting the blame away from the Tricaster LiveKey.

    So, what can be done? Well, NewTek could take a serious look at what I consider to be a serious bug. Most importantly, the alpha matte should be properly aligned. The effect of 422 chroma subsampling (and to some degree chromatic aberration) can be mostly corrected by applying a very slight directional blur to just the chroma channel (it won’t make the video look blurry since the luma isn’t touched).

    In our case, we couldn’t wait any more for NewTek to fix this, so we’ve purchased our first of three Ultimatte 12 keyers. I’m happy to report that this finally solved all our issues with ugly lines, dark or white.

    Attached is an example of how the miss-aligned alpha matte in the Tricaster causes an outline on the right side only. I should point out that we shot this for the purpose of demonstrating the issue, and that I’m not submitting it to be judged by the “Chroma keying technique” judges residing on this forum ;-)

    Click image for larger version. 

Name:	FFC03257-7E50-47C2-802D-70D4D4FBD820.png 
Views:	24 
Size:	481.3 KB 
ID:	142226

  8. #23
    'the write stuff' SBowie's Avatar
    Join Date
    Feb 2003
    Location
    The stars at night are big and bright
    Posts
    18,676
    4:2:2 is pretty commonplace, as I'm sure you know. 4:2:0 too, for that matter, which is a bigger issue. I would be appreciative, anyway, if you could send me the case number(s) for your bug reports. Since I'm having lunch tomorrow with the gentleman who wrote our keyer, it might be opportune.
    --
    Regards, Steve
    Forum Moderator
    ("You've got to ask yourself one question ... 'Do I feel lucky?' Well, do ya, spammer?")

  9. #24
    I wonder if LiveMatte Pro in the next release of TC1/VMC1 will alter any of this.
    Kane Peterson
    Key Accounts Sales Engineer
    NewTek, Inc.

  10. #25
    Registered User
    Join Date
    Sep 2017
    Location
    Sweden
    Posts
    6
    Quote Originally Posted by SBowie View Post
    4:2:2 is pretty commonplace, as I'm sure you know. 4:2:0 too, for that matter, which is a bigger issue. I would be appreciative, anyway, if you could send me the case number(s) for your bug reports. Since I'm having lunch tomorrow with the gentleman who wrote our keyer, it might be opportune.
    Great, I’ll send you the case numbers privately later today.

    Yes, 422 is currently as good as it gets, at least over single link SDI. That’s why Ross even made a special camera for keying with a separate SDI output for just the chroma signal be able to use the full 444 information in their own keyer. Given that the 422 chroma sampling is a known factor, I think any keyer that aspires to be “Pro” should have functionality to deal with it. That it can be done efficiently is illustrated by how Ultimatte 12 does it. Strangely enough, many professional keying plug-ins don’t have the functionality built in either and expect the user to deal with it separately. Best solution I’ve found for this is the Magic Bullet Deartifactor.


    Here is a link to an uncompressed frame directly from the camera illustrating bright edges over dark backgrounds with ProCyc chroma color (couldn’t upload it here for some reason): https://app.box.com/s/luhzvv6eet2fkiixa7w2z9m8dgzcoaf1

    Keying that image over a black background is a tough job for any keyer, but I’m hoping someone else could verify the alpha shift by comparing their results when keying inside TC and in an external software.

  11. #26
    Registered User
    Join Date
    Sep 2017
    Location
    Sweden
    Posts
    6
    I've prepared some more images to better illustrate the issue. Again, this is intended to illustrate the issue and not proper green screen lighting set-up. I pulled a key in the Tricaster and then tried to match the settings in Keylight, After Effects. On one of the samples I also applied a small chroma blur using the Magic Bullet Deartifactor. As you can see, applying some chroma blur helped get rid of most of the dark halo around the white shirt. Even without chroma blur, there is much less of a dark edge in the Keylight sample. Most importantly, the edge is pread evenly across the subject and not just on the right side. I also attached the original green screen shot if anyone wants to try this themselves.

    Click image for larger version. 

Name:	Lisa_Tricaster_8000.jpg 
Views:	22 
Size:	889.8 KB 
ID:	142231Click image for larger version. 

Name:	Lisa_Keylight_ChromaBlur.jpg 
Views:	25 
Size:	302.3 KB 
ID:	142232Click image for larger version. 

Name:	Lisa_Keylight.jpg 
Views:	22 
Size:	316.7 KB 
ID:	142233Click image for larger version. 

Name:	Lisa_green_screen.jpg 
Views:	23 
Size:	805.3 KB 
ID:	142234

  12. #27
    'the write stuff' SBowie's Avatar
    Join Date
    Feb 2003
    Location
    The stars at night are big and bright
    Posts
    18,676
    Just as a passing comment, there are a lot of keying techniques that can be applied in a renderer that are not possible, or not easily possible, to apply in realtime. Blurs are one of these things.
    --
    Regards, Steve
    Forum Moderator
    ("You've got to ask yourself one question ... 'Do I feel lucky?' Well, do ya, spammer?")

  13. #28
    NewTek Engineering ACross's Avatar
    Join Date
    Jul 2003
    Location
    San Antonio
    Posts
    775
    Quote Originally Posted by SBowie View Post
    Just as a passing comment, there are a lot of keying techniques that can be applied in a renderer that are not possible, or not easily possible, to apply in realtime. Blurs are one of these things.
    I don't want to go to deep into this here, but I consider that blurring is not a good thing to do at all in any keyer and so my choice not to use it is unrelated to CPU or GPU related burden. The reason that a blur is a hack is that you use it to mask the fact that you are failing to deduce a good alpha channel from an image at full resolution (and so you need to blur out the problems); to me this is failure.

    In regards to comments on our keyer in general ... remember that we are doing 16+ channels of UHD at 60fps (along with everything else we do!). This is not the same as what you can do in a post processing tool like AfterEffects that cannot even do a single stream of real-time 1080p60. That said, any imagery that we do not key correctly is always good and we always look at it and use it to improve things.

  14. #29
    Registered User
    Join Date
    Sep 2017
    Location
    Sweden
    Posts
    6
    Quote Originally Posted by SBowie View Post
    Just as a passing comment, there are a lot of keying techniques that can be applied in a renderer that are not possible, or not easily possible, to apply in realtime. Blurs are one of these things.
    Well, vMix has a edge softening blur filter that works very well running on a regular PC. In the high-end segment we have this (still running in real-time on regular PC hardware): https://youtu.be/ceddbBnl95s

    I know that a Gaussian Blur is very computationally expensive, but the chroma blur used here is a simple directional blur that shouldn’t have much impact on performance if it’s implemented correctly.

    I understand that edge feathering can be considered a “hack”, but the fact remains that sometimes it’s the only way to get a working key. The world is rarely perfect, and the same applies to conditions for keying. If a tool tool gets the job done, why not use it, even if it masks the “failure” to deduce a good alpha channel?

    Chroma blur (let’s call it Chroma Correction instead) is not the same thing as blurring a matte to mask a bad green screen. It’s very common practice to apply this correction before the actual keying is done, as a way to correct a universal issue when keying high contrast video material.

    I’m still curious about if the “edge on the right side” issue we originally discussed here is going to be addressed?
    Last edited by Johan Folke; 07-13-2018 at 11:44 AM.

  15. #30
    NewTek Engineering ACross's Avatar
    Join Date
    Jul 2003
    Location
    San Antonio
    Posts
    775
    Quote Originally Posted by Johan Folke View Post
    I’m still curious about if the “edge on the right side” issue we originally discussed here is going to be addressed?
    I am afraid I do not talk discuss future development directions and plans on forums

    I still believe that for most people LiveMatte is way better than most other keyers out there, especially given what the system does.

Page 2 of 3 FirstFirst 123 LastLast

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •