OK, Neil . . . now that you've played with it for long enough . . .
Is it an HD DVX?
Is it an HD DVX?
Finally! CCD specs and DSP released!
- Thread starter Jarred Land
- Start date
BobDiaz
The Professor
Emanuel said:
You are correct, my calculator also says, 1,036,800 unique points for pixels.
So, the T-Shirt should read, "I bought an HD Camera an all I got was 1,036,800 unique points for pixels!"
What can I say, Steve's model was correct and mine was off. :crybaby:
Steve ---> :nads: <--- Bob
Oh well, you can't win them all...
Bob Diaz
Emanuel
Cross-Examiner
:dankk2: Bob!
rogar-Thinking(DBG...in Europe he should pay VAT...:furious3:
Amusing smiles that Jarred gave us to play, BTW fair play Professor! Our dvxuser.com Professor. One of the best last Jarred's recruitments. Lucky guy... You are a value-add...BobDiaz said:
rogar-Thinking(DBG...in Europe he should pay VAT...:furious3:
Last edited:
ullanta
Veteran
dhffhd said:
Well, there's compression in there; that's the rub. You can do a lot more interesting stuff when playing with the raw output of the CCD than when trying to play with DV-compressed footage.
Basically, any modification to the signal can't be undone. So you can't even do as much with an uncompressed image as you can with the raw CCD info; and you can do even less with compressed footage.
The Andromeda approach is sort of what you're talking about - I don't know what format exactly you capture from the thing, but you have to use their fancy-schmancy software to turn it into an image. I don't know if the Andromeda still lets you record to DV tape at the same time... but if it does, that's your dream come true...?
ullanta
Veteran
limo991 said:
No, not lossy, but a better lossless algorithm. Huffman requires precise a-priori knowledge about distribution of data to be effective, and thus requires too many passes through the data to determine optimum data block size and such... Yuck! And the unique prefix requirement is at odds with optimal compression. There's plenty of better ways to losslessly compress!
Neopics
Well-known member
The chroma blockiness is an artifact coming straight from the camera. It can be seen without keying. Although the HVX outputs a 4:2:2 signal, the CCDs are generating an image that's visually somewhere between 4:2:2 and 4:2:0.Neil Rowe said:
But more to the point, people can get good keys despite the HVX's inherent chroma blockiness. A lot of good keying software can now blur the chroma channels to overcome the sampling issues of 4:1:1 and 4:2:0. (This is an issue Philip Williams and I did a fair bit of research on in early 2000.)
BobDiaz
The Professor
Neopics said:
I think you have sait it the best way; the HVX's 4:2:2 sits between a full 4:2:2 and 4:2:0. On the edge of a red card over a greenscreen, one scan line down from the green pixel, there is a red pixel, but not a a green pixel. It is additional color information, but not the full color information.
Bob Diaz
4:2:1 lol.
Justin_Kirch
Runner Up.
haha, guys, i'm sorta late jumping back into this thread. just wanted to say that i just patented my quote, so if you suckas are going to make any bumper stickers or t-shirts, i require atleast 15% of the profit.....
but seriously, with all this technical mumbo jumbo, pixels HAVE to kiss my ass. only because i have no idea what any of this stuff means, lol. i just use the camera, i don't give a crap what the resolution is. it's a matter of how you use the equipment to it's best ability, in my opinion that is.
but seriously, with all this technical mumbo jumbo, pixels HAVE to kiss my ass. only because i have no idea what any of this stuff means, lol. i just use the camera, i don't give a crap what the resolution is. it's a matter of how you use the equipment to it's best ability, in my opinion that is.
Morgan_Rozender
Active member
Is it generally accepted that 4:2:2 (DVCPROHD is superior to 4:2:0 (HDV). And 4:2:0 (HDV) is superior to 4:1:1 (DV) ?
Barry_Green
Moderator
It is generally accepted that 4:2:2 is better than 4:2:0, yes. And 4:2:2 is better than 4:1:1.
But 4:2:0 and 4:1:1 are fairly equivalent to each other. Both reduce chroma to 1/4 the luma res; 4:1:1 does it by giving a new chroma sample every line, and each sample is 4 pixels wide (so 4x1); 4:2:0 gives a new chroma sample every other line, and each sample is 2 pixels wide and 2 pixels tall (2x2).
In interlace mode 4:2:0 is, as Graeme Nattress calls it, "ghastly." In progressive it works well.
You can see comparative examples of 4:2:2 vs. 4:2:0 vs. 4:1:1 at http://www.eleventy.org/dv-en/index.php?actie=resultaat
But 4:2:0 and 4:1:1 are fairly equivalent to each other. Both reduce chroma to 1/4 the luma res; 4:1:1 does it by giving a new chroma sample every line, and each sample is 4 pixels wide (so 4x1); 4:2:0 gives a new chroma sample every other line, and each sample is 2 pixels wide and 2 pixels tall (2x2).
In interlace mode 4:2:0 is, as Graeme Nattress calls it, "ghastly." In progressive it works well.
You can see comparative examples of 4:2:2 vs. 4:2:0 vs. 4:1:1 at http://www.eleventy.org/dv-en/index.php?actie=resultaat
Justin_Kirch
Runner Up.
good link there Barry. it's amazing how bad you can actually see DV pixelation occur in such circumstances.
Yuval Shrem
Active member
BobDiaz said:
A few questions and thoughts with response to the CCD specs:
Is this post by Bob Dias (quoted above) correct?
Will a BLUE SCREEN shoot create a lesser quality of chroma-keying?
I tend to use both green and blue screens (depending on the colors of the foreground subject) and was planning to use blue screen in outdoor scenes for keying purposes.
For a narrative project with a film-out delivery, will I be better off using the XL-H1 (for the additional resolution?
When using HVX200 will there be a noticeable difference between keying Blue-screen footage vs. Green-screen footage?
What you actually see there is "A quality comparison between DV-25Mbps and DV-50Mbps " to quote the website directly. Hence the difference between the posted examples is down to far more than colour space. The 4:2:2 samples benefit from double the bit rate compared to the others, and far lower overall compression. It is not possible to say on the basis of these results how much the improvement is due to 4:2:2, and how much to other factors - like is not being compared with like.Barry_Green said:
If 4:2:2 was employed within a 25 Mbit system, it would lead to greater compression being necessary generally to cope with the increased no of samples (33% more, relative to 4:2:0 and 4:1:1). It would be interesting to see how the relative samples compared then.
I will agree DVC PRO 50 gives superior results to DV25. I don't agree it's all down to colourspace.
harddrive said:
I've done more than a few PAL DVDs from Digibeta (4:2:2), at first I was acquiring as 4:2:2 and chewing up a lot of disk space. Then I tried just using the DV25 via 1394 from the J30s and to be honest one really couldn't pick the difference once it was encoded for DVD. I'd imagine though that if the comparison was starting from NSTCs 4:1:1 then the difference would be noticeable.
I wasn't doing anything to the footage though, all editing, FXs etc were done at 4:2:2
David Jimerson said:
well, yes...and no. at a glance, it surpasses the DVX in terms of lattitude and dynamic range, and it has SO MANY MORE features...some hidden like the better signal processing and such. but the P2 workflow is by far the coolest thing. its definitely HD though. simply crazy definition compared to the DVX. plus the longer zoom, and well, i could go on for a while. but ill just say that i love the thing and if youve shot DVX youll feel right at home with it, but youll also feel like youve just stepped up to a much higher level, and opened up a whole new world of seemingly endless possibilities with the cam. i cant say that ive even swithced the thing into any SD mode once since ive had it.. and tape? hahaha.. yeah right. that would be like running a race car off of butter. what can i say. its a little dream cam with an affordable price tag. if you want to make motion pictures that can stand up image wise in the theatre then this thing can take you there. im sure i sound a little excited, but quite honestly.. i am.
color sampling models
color sampling models
Here is the output of a matlab model of color sampling. It is for UNCOMPRESSED since it does not simulate compression. The models involve smoothed chroma. So, this is the theoretical maximum capability of color sampling. The model was run on a PAL test chart. This chart has color bars with single and double pixel lines which really show what we loose in chroma with sampling. This model represents progressive quality.
4:4:4 http://img457.imageshack.us/img457/1424/palcolorchart4449xl.png
4:2:2 http://img528.imageshack.us/img528/8349/palcolorchart4222vr.png
4:2:0 http://img457.imageshack.us/img457/378/palcolorchart4202sd.png
4:1:1 http://img149.imageshack.us/img149/7529/palcolorchart4117jd.png
The actual data size is:
4:4:4 100%, 4:2:2 67%, 4:2:0 and 4:1:1 50%. This practically means that a codec of same compression (quality) would have to use 2x the data rate to jump from 4:1:1 to 4:4:4.
color sampling models
Here is the output of a matlab model of color sampling. It is for UNCOMPRESSED since it does not simulate compression. The models involve smoothed chroma. So, this is the theoretical maximum capability of color sampling. The model was run on a PAL test chart. This chart has color bars with single and double pixel lines which really show what we loose in chroma with sampling. This model represents progressive quality.
4:4:4 http://img457.imageshack.us/img457/1424/palcolorchart4449xl.png
4:2:2 http://img528.imageshack.us/img528/8349/palcolorchart4222vr.png
4:2:0 http://img457.imageshack.us/img457/378/palcolorchart4202sd.png
4:1:1 http://img149.imageshack.us/img149/7529/palcolorchart4117jd.png
The actual data size is:
4:4:4 100%, 4:2:2 67%, 4:2:0 and 4:1:1 50%. This practically means that a codec of same compression (quality) would have to use 2x the data rate to jump from 4:1:1 to 4:4:4.