#3DS stars
33 messages in this thread
Is there ba routine out there that can generated stars aka Star Trek? I have
the star generator that came with 3DS R2 but they don't fly past the camera,
they act as a background. I'm trying to fly through them.
Any suggestions would be helpful.
You would have to be at several thousand times the speed of the light before
noticing that kind of movement. And, assuming you managed to get up to that
speed, you would be in total darkness.
Hi Gus,
Your name was suggested to me by one of the 3DS junkies…maybe you can shed
light on a problem I'm trying to crack.
I have shot some footage of a man standing against a *Blue Screen* and on a
Blue Floor…I want to Chromakey out the screen and floor and composite the man
into a 3DS scene. I have successfully stored the 300 frames of video footage
and now need a routine that'll chromakey. Since I can't find one thats cheaply
available commercially I'm gonna write one myself *BUT* I need to figure out a
routine, that will in essence, build an Alpha channel based on the blueness of
the frame. (i.e shadows are retained but raw background isn't).
Do you know of a good algorithm for processing the incoming pixels??
Just transferring the blue channel to the alpha channel gives pretty crappy
results. Using Adobe Premiere gives acceptable results *BUT* sooo slow and it
can't cope with TGA's and its real stupid about batch frame processing..so the
only solution is D-I-Y …please help.
SK.
I might have just what you want. I have an IPAS that will do the filtering you
need. As long as you do everything within Video Post, you're done. Let Greg
fill you in with the 3D Studio side of it and if needed, I can explain the
technical aspects.
Suresh:
We have completed the first portion of the exact IPAS routine that you are
talking about!
There is just a bit more work to get the shadows looking good on the "new"
background, by making sure that they have the proper amount of alpha.
It's a fun algorithm! (Isn't it, Gus? <g>)
Should be complete in a few days, we'll let you know!
Greg Pyros
This sounds like a great (and incredibly useful) IPAS routine, Greg & Gus.
Thank you very much for making it happen… I'm looking forward to seeing some
demo video from you (Greg) showing it off!
– G
Wow. That sounds like a great IPAS.
John Tissavary (La Luna cie)
I brought this here so other people can use the information instead of keeping
it in private mail. For that, people pay me money. Here, it's just a jolly
season I guess <g>…
You are having problems because you are dealing with a RGB color model. You are
trying to separate a chroma value. For that, you need to convert the color
model from RGB to something that gives you chroma and luminance. HSL for
instance (YUV also works but it's trickier to get the color right – but the Y
component gives a cleaner luminance — oh well…). Once converted, you can
isolate the chroma value and the luminance determines the amount of
"transparency" to apply (that's what gets saved out to an "alpha" file). In my
case, I also have control over "deviations" both in chroma and in luma. This
allows some margin for not so blue blues and contamination by warmer light and
stuff like that.
If you don't know how to convert color models, I won't attempt to explain here
as "that" could be a little complicated if not too extensive. If that's the
case, I would suggest a couple of books. Let me know if you need these titles
and I can go find them…
>> in reply…
All right, here they go…
Computer Graphics Principle and Practice James Foley, Andries vanDam, Steven
Feiner, and John Hughes Published by Addison-Wesley Publishing Company ISBN
0-201-12110-7
Digital Image Processing Rafael C Gonzales and Richard E Woods Published by
Addison-Wesley Publishing Company ISBN 0-201-50803-6
Digital Image Processing William K Pratt Published by Wiley-Interscience ISBN
0-471-85766-1
>> One more question, is YIQ the same as YUV??
Only if you think of them as two color models. YIQ represents Luminance (Y),
Hue (I), and Chrominance (Q). This is the same as the HLS I was talking about
before. YUV is a TV transmission color model where the U and the V signals are
related to the I and Q signals by a simple rotation of coordinates in color
space.
Gus, I thought YUV refered to the betacam components Y, R-Y, B-Y FOR PAL and
YIQ refered to the same betacam components for NTSC. A slightly shorter
alphabet soup of stating the encoding of the components from RGB Therefore,
when correctly used if I heard the reference YUV it pertained to the signal
from a component out of a PAL betacam deck (or any deck using that encoding
method for component PAL)
Perhaps YUV ties into the fact that CCIR-601 was developed to be compatible
with NTSC and PAL, and that's why it's used in the Abekas. I'm really
confused, though, since I thought YUV was just another way to encode light and
color info, and I didn't know or suspect it was related to a broadcast signal
format.
John Tissavary (La Luna cie)
You are right. I don't remember saying otherwise though. The YUV coordinate
system was initially proposed as the NTSC transmission standard, but it was
later replaced by the YIQ system because it was found that the I and Q signals
could be reduced in bandwidth to a greater degree than the U and V signals for
an equal level of visual quality. Come to think of it, what am I saying…
Aren't you the TV wizard?
At any rate, this is what I was saying when I was talking about the virtues of
D1 being a bit overstated. By the time you get the image on tape, it has been
reduced quite a bit from its original RGB form. Technically, I can compare both
D1 and JPEG as lossy compression. The differences are obvious and worlds appart
but people shouldn't discard "digital" alternatives just because they have
"JPEG" attached to it. It bugs me when people talk about D1 as the best
"quality" as in picture definition. D1's virtues are in being digital and as
such, infinitely (theoretically) reproducible and suffering no loss in the
transport process. As far as image quality goes, or amount of information
present in order to reproduce an image, it has less than what something like a
PAR unit would display. Or a laser disk for that matter. Heck, even less than
Hi8. The problem is that a tremendous amount of information sent to an analog
deck as a Hi8 or a LVR is lost on the way in and on the way out of it. And to
top it all off, once you reach the limit in bandwidth of what can be
broadcasted without violating FCC's regulation, you're done. I can't wait to
see the day we find ourselves free of NTSC and go all digital. Can you imagine
the kinds of things people will be bragging about? Thing is, you've got D2 and
D1. What's next, D -1, D -2, etc? <g>
Oh well, as usual I went way off again…
Gus,
>>As far as image quality goes, or amount of information present in order to
reproduce an image, it has less than what something like a PAR unit would
display. Or a laser disk for that matter.
Well Gus .. I guess you expected you'd get some mail on that one <g>
As far as I understand (and I do stand to be corrected .. if appropriate) JPEG,
like D1 is 4:2:2. The difference being that JPEG is further compressed (thru a
Fourier Transform based algo) to produce a further reduction in bandwidth.
Thus, D1 is the best that JPEG at "0" q factor can duplicate. Now the PAR
system can bare sustain 3 MB/sec thruput which equate to about 18-20 q-factor.
The MAX system (using parrallel SCSI HDD's) can sustain 7 MB/sec .. about 7-9
q-factor … which even for a still image is visually lossless. Now at that
point, we are equivalent to D1.
I do agree that a laser disk will have higher quality that all being discussed.
Again, if I've made an error is arguement, please feel free to "slam" away <g>
Ang
I won't argue anything you said. But I got your attention, didn't I? <g>
You did not disagree with the point I was trying to make. I said JPEG and D1
are worlds appart (the $148,000 difference should make that pretty clear. <g>).
My point was to emphasize the fact both methods are digital and can't be
compared with analog stuff. You can't compare D1 to Hi8 in the same level as
you would compare a LVR to Hi8.
When I said D1 has less quantity of information encoded, that's based in
theory, not practice. That's because JPEG at 0 doesn't go through the transform
you mentioned (now, go find a system able to pump that out)…
Gus,
>>JPEG at 0 doesn't go through the transform you mentioned (now, go find a
system able to pump that out)…
Well now … that's the 1000 lb gorilla .. isn't it. Ya'know, there is more BS
on the ad's in mag's (and in recent product lit) regarding motion JPEG
solutions and their capabilities. When I call, it takes all my effort to pin
people down on EXACTLY what their systems are capable of.
A good recent example is the MegaMotion board by Alpha Systems (you
may have caught thier ad's in New Media).
They say " … professional video production to your PC with Broadcast
Quality (CCIR-601, 4:2:2 YUV) .."
Well, I gave them a call to see what was going on. Well they can take
as an incomming signal CCIR 601 … but when they digitize its only
320×200! … WHAT CRAP. The other thing I am seeing is that they
will say something like…
".. full motion, 60 fields a second, and full screen video
capable…"
What is not said is that you can't have them at the same time!
Since the MAX system (since its based on SCSI) will allow you to
parrallel to disks, you can actually get the JPEG down to "visual
lossless" levels (JPEG <10). I think in the next itteration they
might add features so that the board will look like an Abekus Digital
Disk Recorder (ie 422 controllable) … that would give you the
possibility of a system that would do what the Abekus can do … but
almost unlimited duration (not really but you will have 7 SCSI device
ports for large HDD's).
Ang
He, he, he… I've gone through the same. I did called all those
people and the kind of answers I got would fill a comic book. What
really gets distorted is the idea of what is "broadcast" quality. I
guess it means that if you can discern an apple from an orange, it's
got passed the "broadcast" test! <g>
Here's the funny part, though. I and a whole bunch of folks over at
FTI have played a good bit with the Abekas (D1) and one of their many
LVR's, and everyone who looks at the output agrees that the Abekas is
better. Why is this? And if it's a subjective thing, why is it
unanimous?
John Tissavary (La Luna cie)
>> Why is this? And if it's a subjective thing, why is it unanimous?
A few years back I attended a one year program to learn how to do
scientific data evaluation, setting up statistical arrays and such.
It was for marine science but the principals apply universally. I
later became an instructor in this program. Many times when I hear
reports (results) such as yours the first thing I question is the
test methods that were in place because it is just about impossible
to get unanimous results in a truly scientificly conducted test no
matter how ridiculously obvious the comparison is. How many
placebos? What were the controls? How many samples were done? How
many trials per sample? A simple test like this is generally
objective if the test was done with test instruments monitoring
certain attributes of the two signals against a measured scale. If no
scale was used then it was probably subjective. If it was subjective
then how were the opinions suaded? Were the individuals isolated from
each other so as not to allow peer persuasion? Were the individuals
allowed to know what they were looking at? ie which monitor was LVR
and which was D1 at the time of decision. I assume the program was
identical on each and the monitors were identical or the same with a
switch. The point I'm trying to make is that a true test is really
difficult to set up properly and if no attempt was made to do so the
results are meaningless. Finally, outside of what I learned in this
program, if something is just plain obvious then unanimous results
would be possible, but I doubt the difference in LVR and D1 is that
obvious. Many times we see what we want and hope to see.
Don,
>> what is the test setup
The most common example of how to see EXACTLY what you are getting at
is to walk in the AV dept. of any Sears, Ward, etc … All the TV's
are playing the same picture (sometimes), but each is setup "not to
the max" .. and many times a picture "looks" better on a TV that is
1/2 the price.
The tests weren't really tests, they were just opinions based on
seeing output from both. All of us were biased in some way, and the
animations weren't always the same, although they were very similar,
simple forensic (simulation) animations. I'm sure under test
conditions there wouldn't be a unanimous opinion, but I'm definitely
partial to D1, subjective as it may be. The main difference to me
is that LVR output seems a little colder than D1, but it's personal
opinion/preference. And I wouldn't be surprised if under true test
conditions I wouldn't pick the LVR as better under certain
conditions/animations. It's all voodoo to me anyway.
John Tissavary (La Luna cie)
No offense to this but most people run "tests" that are not really
valid but design the tests to support their prejudices, even
scientists. My training taught me to realize the importance of
proper testing and recognize when published statements are just based
on bias or really bonafied results of a properly conducted test
whether it be subjective or objective testing.
In addition to what Don said, I'd be curious to find out what the
program was that was viewed. Computer animation? At what resolution
was the LVR recorded at? What type of input? What graphics card?
First generation on an LVR at a high resolution from a good graphics
card and encoder should be very difficult to tell from D1. I totally
agree with the power of D1 for generational stability, etc., but I
also wonder how the test was conducted to compare a $17,000 unit with
a $115,000 unit. It is well known that an AT-Vista can give you
twice the horizontal resolution of D1, and in some tests this would
become apparent.
I recall a technical message Gus Grubba left describing how the
conversion of RGB to YUV 4:2:2 gives you less bandwidth for the data,
and that he considered this is a form of lossy compression.
Not knocking D1, just the "scientific testing" procedure!
There was no test, especially not a scientific one (you know me better
than that <g>!!!). It was just a group of totally biased,
uncontrolled, overeducated, undersexed, socially inept, morally
corrupt, people like me <g>. As I said in a reply to Don L.: I'm
sure under other circumstances, in a blind taste test I'd pick the
LVR on occasion, too. I did like Abekas playback over the LVR in
most cases I saw at FTI, but I have no idea what resolution they
recorded at. I'm sure it was 720×486, which would hamper LVR a bit.
They don't use PC's, and I'm not sure what kind of video card they've
got in their multi-million dollar roomfull of SGI's. A good one,
probably.
The real price contest is not between the LVR and D1 (no contest
there!), but between an Abekas A60 series DDR and the LVR. I'd pick
the Abekas, but not just because of video quality. I actually
haven't recorded to D1 yet because the edit suite rentals are so
expensive, I generally dump to BetaSP. I'm about to lock down a job
that will probably go D1, though.
how much is an A60 these days and what's the capacity? I was thinking
of getting a portable Abekas to connect to my laptop, you know, for
doodling when I'm on the road <g>. I just saw the prices of Abekas
systems had gone down, I just don't know to what level.
by the traditional logic (especially dealers logic <g,d&r>) of "you
get what you pay for", D1 must be better because it costs a lot more,
right? Actually, I'd just look at what the industry (network TV) CGI
heavy hitters are using (Ron Thornton, Spermquest, 500 nations etc),
and it seems to be the Abekas/D1 route, possibly with an LVR for
high-quality previews. You have that exact capability also, right?
Yes, we do. However, that wasn't the direction that the original
question was taking. On first generation, especially at the higher
horizontal resolution the LVR is capable of, there are times it would
be very hard to distinguish from D1, and may even look better!
However, I totally agree that the "0" generation loss quickly makes
up for these in editing!
We do all our previews to the DPS-PAR, next iteration to Laser Disk,
then final output to D1 if required. (I don't have a D1 in-house
[yet])
>>Aren't you the TV wizard?
Depends on who is judging. <g>
I agree what you said about this D1 thing being over rated. Now, if
you need to take the video material 10-15 generations through
layering then it does have its place. However if the animation is
going to be viewed after only 5 -6 generations then betacam is very
strong. 99% of all my animation renderings are viewed at 3-4
generations, the last being VHS. 3/4 Umatic works fine but I sure
would like to have component editing for all my cuts work. That is
why I'm looking at moving to Beta SP here. If it weren't a dying
format I would consider 1" instead.
YUV is a color space and you can use these formulas to translate to
RGB: Y= 0.30 R + 0.59 G + 0.11 B U= 0.62 R – 0.52 G – 0.10 B V=-0.15
R – 0.29 G + 0.44 B Pierre
Bob,
if you have the Yost Group's particle systems IPAS routines you can use rain
for this effect. You just point the camera upwards and the stars/rain will come
towards you. You could make adjustments to the rain particles shape to make it
look like stars and maybe use glow if you have it.
Oohh. I like that one. Too bad I don't have RAIN IPAS or a need for stars
flying through space.
John Tissavary (La Luna cie)
Welcome back, Martin. I like that stars trick. I'm going to file it for
future reference. Thanks.
Don,
It's nice to be back on the forum, Don, thanks. The stars trick is probably
best credited to Gary Yost, I think, not to me. He has mentioned it once or
twice.
I don't know of a quick and easy way to do fly through stars, but this'll work
with a little patience:
Use stars.ixp or any other way you want to create a series of images (tga, gif,
jpeg, whatever). The more of these you create (all different starfields, of
course), the longer you'll be able to "fly" through them.
Use the images to create star materials. To do this use the image file both as
a texture map and as an opacity map, with the opacity slider set all the way to
100. To make the planes just creat a QUAD in the 2D shaper, then in the 3D
Editor do CREATE/OBJECT/GET SHAPE and import as many of these planes as you
need. They will have normals facing only one way, so if you plan on "turning
around" you'll want to make your material (stars) two sided. Then apply the
matierial with planar mapping, and place the planes one after another, with
enough space in between to give you a feeling of "largness".
The only drawbacks to this technique I can think of are rendering times
(opacity maps take a while to render), and you'll have to experiment to get the
look you want.
Good luck,
John Tissavary (La Luna cie)
If you have Animator Pro, there is a POCO program called STARFIEL.POC that does
this. It will create a flic, any length, up to 2000 stars comin' atcha.