Greenscreen stills are up

[im.thatoneguy]

Insert obligatory keying test:

Obviously no integration or compositing work, just a key. And what a beautiful key it was. Two clicks. The matte did need .8px of gausian filtration because of some de-mosaic artifacts though.

Given time, you could extract a perfect key. Then again, this is just about the simplest green screen test one could conjur up. David Stump said you guys shotmotion and fine hair... will they be posting that on sunday?

[boothba]

lovely.

[Mr. Blonde]

F*** that's clean.

[Jaime Valles]

Unreal. Absolutely unreal, and this isn't even the prototype finished camera or software!

[jbeale]

very nice! it's fun to note also that one of the green Macbeth chart boxes keyed right along with the background.

[jbeale]

I'm curious about something- right after David Stump visited Red and before any of the Red frames went up, someone (Jarred?) posted a very low-res photo of the greenscreen setup- for all I know taken by a cell phone- but the green background was very vivid. Now, looking at the green screen images from Stump's session, the background is very different. Did the lighting change that much? Look at the comparative values of blue, in particular on background:
R,G,B values Left: (Jarred?): (65, 237, 1) Right: (Stump): (7, 224, 161)

The colors on the charts look reasonably OK on the Red test, so it doesn't seem possible that the background should be so different, unless there was a tremendously different lighting setup- like about 5 stops more blue on the background- when the right hand photo was taken.

[Jannard]

The "snap" is not color accurate (much too warm- look at the seeing eye) and has been saturated a ton. Stump's footage is out of our camera and displayed "flat". I would NOT make any opinions based on this.

Jim

[jbeale]

I have heard about a "color matrix" in digital imaging systems, although I can't say I fully understand it. Is this something that happens in the Red camera somewhere between sensor-in and data-out? If so, does the color matrix have final parameters yet, or is it still in process?

Edit: To my surprise, I was actually able to manipulate the Red image at right to approximate the colors of the "snap", by using only Photoshop global curves & tweeks (not isolating any one color). So (in my limited understanding) it seems you could have this much difference just via a color matrix, and contrast curves. Interesting.

[Jannard]

A sensor would ideally spit out a perfect D50 color balance. Since they don't ever hit it exactly, a color matrix is designed to "adjust" for the difference. Ours is a very mild correction.

Jim

[mikkowilson]

Damn that's some clean keying!


- Mikko ... is mirroring files now.