Greenscreen Test
 

Hi guys,

I want to share my test for a greenscreen project with HD100 in studio.

I test the JVC for a solid and safe workflow to make a greenscreen shot, matchmoving and 2.5d matte camera projection.

The studio was 14m x 12m and the file captured from tape (HDV).

In attachment 2 pictures and here the video:
http://www.andreatoniolo.com/sucks/MATCHHD.mov
ATTENTION the matte behind the actor are NOT made by me it's only a reference because I cant publish this month my matte (commercial thing).
The video are matchmoved and than RAW projected in maya (obviously only as sample to trust the matchmoving).

Anyway I can write some of my personal technical advice. :)

That is sourced from the tape? I like it and would like to know the details.

Thank you for your post Andrea. Please give us more details. Was the background set shot with the HD100 too? Tell us about lighting, editing software, shot distances, etc.

Very nice results indeed. What scene-file did you used?

Cool looking shot and nice matchmove test. Looks very tight.
What matchmove software did you use.

So this is just a test matte?
Seemed like the camera projection was breaking up a lot as the camera moved forward. The bottom of the ladder rips off, the window on the left gets sliced and the objects catching the posts seem to also be getting some of the distant background pixels.

The comp could use a bit of edge blending and maybe even light wrap. Black levels on the foreground seem a bit dark.
What's the purple shadow there behind the left ear?

Again, cool shot and nice matchmove!

Hi again,

I try to explain as much as I can.

first of all the shot was not made to match that matte painting (the light is totally wrong and I not used the lightwrap) and the camera projection is really raw bacause the image are not divide by layer but it's a unique imageplan projected with raw polygon models (this cause the ghost effects and the disrup of image).
As you know its impossible to create a matte painting without the plate with the polygon reference (because of the difference in the vanishing point ecc ecc).
My good friend "GINUZ" (a director and compositor from Italy) has try to made the alpha from the test file and has his own opinion about that, probabilly he can write something here.


For sure these are my personal opinion :)
STUDIO
We used a 2 piece of papers "54" merged with a "hot glue" (about 12m by 14m), with 2 iff 1250 + 2 2000w (fresnel) + 4 ianiro 800w and 1 Apollo 5000w.
My DOP Giuliano Rancan say that the camera response good in the green channel.
The actor was 5m from the greenscreen.


POINT
I test 3 difference type of POST-IT and the best was the POST-IT color "green" for matchmoved and also for greenscreen. Avoid the yellow one.


CAMERA
I used the Paolo Ciccone V2 but I test a lot of difference scene files with 5 different people (blond hair and brown hair).
What I think it's that the best it's to mantain the master black and gamma at -2 because if you go to "normal" the green washed out.
I have some problems with SSE and some drops from the ProHD tape; the sse was calibrated and calibrate but still remain in certain situation, in the greenscreen its NOT a problem and all go nice.


DETAILS
I set details from "min" to "off". Min its really good. OFF it's good if you need only to film a silhouette (MTV style) and to put in a video. If you used "OFF" the matte it's really "perfect" but details go off and it's impossible to merge the "pixel" of the HD100 to a "pixel" of matte painting.
Also the hair disappear.
-MIN I think it's the best way.
I think the most difficult challenge it's to merge the "type of pixel" from camera and the "type of pixel" of the object behind (if you try the sharpness filter only noise appear).
(Do not misinterpret me, I like the strange pixel, in the 2003 I made this with a little JVC http://www.andreatoniolo.com/test/fashionist/ )


CHROMATIC ABERRATION
The aberration in the right part it's a big problem because made about 2-4 pixel inside outside the character.
As Nick Jushchyshyn say the purple it's a challenge to delete (because the left part go right).
Probabilly we need to separate with 2 different mask.


LENS DISTORTION
This is another problem. As you see in the matchmoving sample there are a little problem in the right part because of no-lens compensation.
I used the WIDE ANGLE ADAPTER from Fujinon x82 and I try to matchmoved.
I explain the scene:
I start to moved the dolly before the actor enter the scene and in this case you perceive that the scene it's not right matchmoved.
This is not totally right. The scene work good and the matte moved first as I expect but the right part influence the other and in the long time appear that the scene move fast than the actor.
I than try the LENS COMPENSATION in AE and I put to -20. Matchmoved well.
I think that -10 its the best way in certain case.

TYPE OF FILE
I personal recommend to use the raw .M2T file and to convert in a sequence of TGA or TIF.

IN THE END
For me is possible to made good-to-great greenscreen in a studio with 14/16kw and to rec in tape via hdv.
But you need to have the actor always in focus because the out of focus and HDV make some problems.
I think that the shot to greenscreen are only a (important) part of the process and in the process you can resolve in some way the problems that the LENS and HDV can make.

This is my 2 cents :)
cya later

Hello there, long time since I've posted on this wonderful board!

Very interesting indeed, just a quick question:

are you tracking a matt and then applying a projecton map into a 3d model (that approximates the matte scene) then using the matchmove data to that low poly model?

Anway great test!

(p.s I did notice a slight projection mapp overlap error, very slight, but it shears the latter in two parts, i'm assuming its not caused by the QT codec, but the way the projection map + underlying 3d object is made.)

Anhar