CompuServe Messages

#Super Black

    27-Oct-94 22:48:26
Sb: #131059-#Super Black
Fm: Robb Harriss 73023,134
To: Don Landis 71673,3612
<<Alpha is about keying???? I don't think so.>> it sure is in video post production. The alpha channel gives the video switcher the information it needs to make a key hole which is then filled with whatever you specify. You decide if the alpha information creates a solid key or a partial key (transparency). A good linear keyer in a good switcher can take the alpha information and interpret it into a clean key, whether it comes from a character generator, a graphics system, a DVE or even off of tape. In the real world most keys (and their alpha information) are solid rather than transparent. The transparency is what causes trouble when the switcher cannot read the alpha information well. Then you get solids when you shouldn't, or no keying at all, and that results in a crummy inage. <<we define super black differently than the way 3DS does.>> so who's right in their definition, the new guy (3DS) or the industry that's been around 30 years longer? <<In video we would expect super black to take the 0,0,0 black to 0 IRE for the key>> actually, I'd prefer Superblack to be -10 to -40 IRE, we get a much better key that way. It works best with a difference of about 40 ire between the black and the lowest level in the image itself. <<What 3DS super black does is create a wider spread between RENDERED black objects and the 0,0,0 black background in the scene. This spread can be specified in the 3DS set file and has a default of 15, I recall.>> Uh, Huh. That's what it claims in the manual, but what setting creates what level of separation. We've fooled with all sorts of setting and still can't get much if any change. If it needs to be up around 200 then I really don't see the point. <<The only way I know of to generate a value of black below 7.5 IRE is to adjust it in a procamp with the pedestal or black level set-up before feeding the alpha to the key input on your switcher.>> That's work only if you're feeding a pure alpha signal, which doesn't need it since it is made up of pure blacks and whites, and shades of gray to give transparency. If you drop the pedestal on a color image you''ll pull the video level down as well. Pulling the video level up again will change the contrast, and the look of what one hopes is a carefully made image, besides, dropping the pedestal will not create a difference between lowest black and the next lowest level in the image. <<Have you ever done any animation using the alpha key? >> Lots, Thousands over the years, sometimes hundreds in one show, so it's very important to us. <<Seems to me you would require two frame accurate VCR's and a third VCR for the background video. >> That's how it's done and that's the rub, you have to pay for an extra machine in the session and you have to have a switcher that can handle keying information from a tape. Basically you have one tape with the color image (key fill), another with the key information (key source) and one or more tapes with other background images. The edit controller will synchronize it all to the Master deck. A traveling matt (or key in video terms) is an exact replica of the color image to be keyed, but it is all blacks and whites matched to the color image pixel by pixel. On the switcher, the key source is selected as a tape input (say deck C) on the Key Buss. The key fill is selected as the color image on tape (deck A) on the Key Buss. On one of the main ME (mix effects) busses a source tape (B) is selected, and this becomes the background when we set the switcher to Key over tape B. Of course if we're really clever we could be dissolving or wiping between decks B and D at the same time that we're keying (and inserting a DVE move from deck E over all this) as well as a downstream key from the CG all before a master fade to black. All of which it gets hairy in the online room. OK, so we have the traveling matt. It is an exact B&W copy of our animation. Since it matches it will be in the same position and be the same size and shape as the color image. The B&W punches the hole in the background video and the color animation fills that hole exactly. Now it's expensive enough and complicated enough that we like to try to go a different route which when it works saves the cost of a deck and a lot of set up. A luminance key works when there is a great enough difference between standard black (7.5ire) and the next level in the image. A pure white CG is a classic example. It can be keyed all by itself because there is a good 90 ire of difference between the black background and the white foreground. The same can be done with color images. However, you must set the lowest luminace in the color image to no lower than 30 or 40 ire (depending on the switcher) on a 7.5 ire background. That can be very difficult with some images. The trick then is to artificially lower the black in the background, to 00 or even lower. Then the difference between that black and the next level is even greater and a better key is possible. Considering the explanation of Superblack in 3DS I expect it to work as a lower end buffer for the dark regions in my image limiting to a specified level (20, 30, 40ire) whatever I set. The the resulting color image should have a great enough luminence difference to key without an alpha matt (and saving the cost of an extra machine in online). Some of the 3DS images work and some don't. The alpha information in 3DS works perfectly, but a traveling matt right now requires rending the entire animation as an alpha split to disk. We're set up (and have been for years) to go directly to component tape. I don't want to go to disk unless I have to, it just takes up valuable disk space and isn't necessary. I certainly don't mind putting the alpha files to disk while I'm rendering color to tape but I don't want to be writing a deleting thousands of unwanted files. I just want to do Alpha split and tell 3DS to put only the alphas on disk. From my point of view it seems like it should be easy enough, but I'm not a programmer.