Mmmm.. I'm not sure if I can ignore pixels count on a digital image.. that is like saying that a 300x300 4:4:4 image is better than a 1920x1080 4:2:2
I think I have to try it to have a better idea of what is really the best thing. This is only numbers :-)
________________________________
From: owner-ds(at)Softimage.COM on behalf of Jef Huey
Sent: Mon 8/1/2005 2:15 PM
To: DS(at)Softimage.COM
Subject: Re: Varicam or HD
Victor,
I agree - more pixels = more resolution. BUT ...
What we are also talking about in a discussion of 4:2:2 vs 3:1:1 is how
accurately each pixel describes the image in front of the camera.
In 4:2:2 there is a greater COLOR resolution in each pixel. More
subtlety. And that is the "key" (pun intended) to keying. Ignore number
of pixels on a line for the time being. That is picture size. Think
about the quality of the description that each pixel gives you when it
comes time to do a key.
Do you follow me there?
Jef
Victor Wolansky wrote:
> Well what we can tell from the numbers is that seems that the HDCAM
> and VARICAM have the same horizontal color information, 480 pixels,
> but HDCAM have more luma information that VARICAM, 1440.
>
> What I can tell by my little research now is that if you can not
> digitize HDCAM SR, and want something better than HDCAM or VARICAM,
> then you can connect you HD-SDI output of the camera to a Deck like a
> D5, D6, or HDCAM-SR in which case I will have a full sampled 1920:960:960
>
> That is my understanding but I could not find yet information to
> confirm this. Any one knows?
>
> I don't think we are comparing apples and oranges, it is an absolute
> thing, more pixels.. more information.. the keyer will have more info
> to do his job. If we compare format to format, VARICAM have more color
> info per pixel, but less total information per line, what can be the
> bad thing about having more horizontal luma information, I think that
> can't be bad.
>
> Just my opinion.. I can be completely wrong...
>
> Victor Wolansky
>
> DS VFX Artist
>
> WEBsite <http://www.e3post.com/>
>
> Demo Reel <http://www.e3post.com/reels/victorreel.html>
>
> 815 Slaters Lane
>
> Alexandria. VA.
>
> 443-797-3507
>
> -----Original Message-----
> From: owner-ds(at)Softimage.COM [mailto:owner-ds(at)Softimage.COM] On Behalf
> Of swaffordt(at)bellsouth.net
> Sent: Monday, August 01, 2005 8:42 AM
> To: DS(at)Softimage.COM
> Subject: RE: Varicam or HD
>
> Jeff... great information, thanks... btw, your reel rocks!
>
> TBONE
>
> -----Original Message-----
>
> From: owner-ds(at)Softimage.COM [mailto:owner-ds(at)Softimage.COM] On Behalf Of
>
> Jef Huey
>
> Sent: Monday, August 01, 2005 10:26 AM
>
> To: DS(at)Softimage.COM
>
> Subject: Re: Varicam or HD
>
> My gut instinct about this analysis is that it has its own flaws.
>
> It seems to be one of comparing apple and oranges.
>
> Alot of the numbers are base on totals per line. I think that is not
>
> valid when it comes to blue screen work. What matters there is color
>
> resolution of each Pixel, not the total for a line. The keyer must make
>
> decisions about what is "Green" and what is not. If color of each
>
> pixel is sampled at a higher resolution (4:2:2), then you will be
>
> better off. That is why 4:4:4 RGB keys are so desirable.
>
> At least that is how I am reading it and I definitely do NOT have a math
>
> head.
>
> Where did this info come from - Sony maybe ;) ?
>
> Jef
>
> Victor Wolansky wrote:
>
>>
>
>> *I found this searching on this topic, I think is very useful
>
>> information to share..*
>
>>
>
>>
>
>> * *
>
>>
>
>>
>
>> * *
>
>>
>
>>
>
>> *Riddle#2: Does the Varicam have more Color Space than the F900?
>
>> Is the Varicam better for green and blue screen matting?*
>
>>
>
>> What I'm discovering is a bit of an eye opener in terms of Color Space
>
>> and Image Sampling between the Sony CineAlta and the Panasonic Varicam.
>
>>
>
>> Have you heard the common misconception, prevalent across the
>
>> industry, that the Varicam records a 4:2:2 color space to tape? The
>
>> argument goes like this ...... while the Varicam has less resolution
>
>> in luma, it has more resolution in chroma compared to the Sony F900
>
>> CineAlta. So this means the colors are sharper, therefore the over all
>
>> image will tend to look relatively sharper as well, despite there
>
>> being less pixels in total compared to the F900. Consequently, this
>
>> helps the Varicam produce an image that looks just as sharp as the
>
>> F900. While the images are indeed amazingly similar, this reasoning is
>
>> bogus.
>
>>
>
>> And if this were true, then the Varicam might turn out to be the
>
>> better tool for doing green and blue screen work. On the face of
>
>> things, it might seem the Varicam will do a better job with pulling
>
>> mattes with double the color resolution. After all you are keying to a
>
>> color, and if the colors are sharper, then it should work better -
>
>> right? Again, this reasoning is bogus. The real story is a little more
>
>> involved, as I will try to explain.
>
>>
>
>> It is amazing how entrenched the misconception is that the Varicam has
>
>> approximately double the color space of the F900. But in a sense this
>
>> is true ..... it all depends on how you look at it! Lets start by
>
>> looking at precisely how each HD camcorder system processes and
>
>> records its images to tape. But in so doing, we can't lose sight of
>
>> the fact that the starting point for each camcorder is different. The
>
>> Sony CineAlta has approximately twice the number of pixels per CCD
>
>> compared to the Panasonic Varicam (2.2 megapixels for the F900 verses
>
>> 1 million pixels for the Varicam).
>
>>
>
>> The first step in both cases is pre-filtering in luma. Both systems
>
>> use the same reduction factor, namely 1.33^* (recurring). The F900
>
>> starts with 1920 pixels per line, and filters down to 1440 in luma
>
>> (1920 divided by 1.33 = 1440). The Varicam starts with 1280 pixels per
>
>> line, and filters down to 960 in luma (1280 divided by 1.33 = 960).
>
>> The original "1920", and the original "1280", are defined by their
>
>> respective format specifications.
>
>>
>
>> Next, chroma is sampled at 3:1:1 in the case of the Sony camcorder,
>
>> and at 4:2:2 in the case of the Varicam camcorder. So the CineAlta has
>
>> 1440 in luma and (1440/3)= 480 in chroma, while the Varicam has 960 in
>
>> luma and (960/2) = 480 in chroma. Note that both systems now have
>
>> precisely the same number of pixels per line, namely 480, representing
>
>> color information!!!
>
>>
>
>> Just to be clear about the mathematics I used here, the divisors 3 and
>
>> 2 respectively, come from the how each camcorder subsamples its color
>
>> space. The CineAlta is 3:1:1, so it subsamples its chroma channels by
>
>> a factor of 3 in relation to its luma channel. Specifically, it
>
>> samples chroma at 1/3 (one third) that of the luma channel. So this is
>
>> why I divided by 3. The luma is sampled at 1440 /3 = 480 which is the
>
>> chroma sampling. Similarly, the Varicam is 4:2:2, so it subsamples its
>
>> chroma channels by a factor of 2 in relation to its luma channel.
>
>> Specifically, it samples chroma at 1/2 (one half) that of the luma
>
>> channel. So this is why I divided by 2. The luma is sampled at 960 /2
>
>> = 480 which is the chroma sampling.
>
>>
>
>> While both systems (coincidentally) have equal chroma resolution, the
>
>> Sony HDCAM has 50% more luma resolution. This is not surprising
>
>> considering it started out with twice the number of pixels in the
>
>> first place.
>
>>
>
>> Another way to look at this is as follows:
>
>>
>
>> HDCAM = 1440 in luma, and (1440 /3) = *480* in chroma = 1440:480:480 =
>
>> 3:1:1
>
>>
>
>> HD Varicam = 960 in luma, and (960 /2) = *480* in chroma = 960:480:480
>
>> = 4:2:2
>
>>
>
>> In an absolute sense, the CineAlta and the Varicam have identical
>
>> color space i.e. *480* pixels per line!!!
>
>>
>
>> But relatively speaking, the Varicam samples a slightly lower
>
>> resolution image at a slightly higher rate for its colors.
>
>> "A-little-more-of-a-little-less" means there is no net gain in terms
>
>> of color space compared to the CineAlta. However, the CineAlta has a
>
>> comparative advantage in terms of resolution in luma (or the black and
>
>> white component of the image) by a factor of a half i.e. 50% more
>
>> resolution in luma. Or 1440 pixels compared to 960 pixels
>
>> (mathematically speaking: 960 x 50% = 480, then 480 + 960 = 1440) (or
>
>> 960 x 150% = 1440) (or 1440/150% = 960) etc.
>
>>
>
>> Sony's 3:1:1 exactly equals Panasonic's 4:2:2 in terms of color space
>
>> in an absolute sense. But in respect to the Varicam's own inherent
>
>> line resolution, it has twice the rate of color sampling compared to
>
>> the CineAlta. This leads to the conundrum that it is true to say the
>
>> CineAlta and the Varicam have identical color space, but at the same
>
>> time, it is also true to say the Varicam has twice the color space of
>
>> the CineAlta. This is of course is contradictory.
>
>>
>
>> The truth of the matter can only be resolved when you look at the
>
>> total number of pixels per line that are sampled for color information.
>
>>
>
>> **3:1:1 color sampling of 50% more line resolution = 4:2:2 color
>
>> sampling of 50% less line resolution.**
>
>>
>
>> To complete the story, both camcorders sample at 10 bit, pre-filter,
>
>> compress, and record to tape at 8 bit. In other words, both systems
>
>> work in a very similar way, and indeed, they both produce very similar
>
>> looking images.
>
>>
>
>> Many people believe that the Varicam performs better for green and
>
>> blue screen matting processes, compared to the Sony F900. I think the
>
>> reality is that they are about the same in their blue and green-screen
>
>> matting capabilities.
>
>>
>
>> Just as an aside, the Sony HDCAM (D-11) compression scheme is based
>
>> upon a 1440 x 1080 image format and 3:1:1 chroma subsampling, as
>
>> described above. Downsampling is required upon compression, and
>
>> upsampling is required upon decompression. The downsampling inherent
>
>> in HDCAM compression of 1080i or 1080p means that luma sample aspect
>
>> ratio is effectively 4/3 (and chroma subsampling is equivalent to
>
>> 4:1:1) relative to the original 1920x1080 image array. So luma is
>
>> multiplied by 4/3 or 1.33^* (recurring) i.e. 1440 multiplied by 1.33^*
>
>> = 1920. Hence it is 3:1:1 color space in respect to 1440 x 1080, but
>
>> really only 4:1:1 color space in respect to the originally captured
>
>> 1920 x 1080, and also the 1920 x 1080 produced after decompression of
>
>> the signal and upsampling. In some ways it is hokus pokus with numbers
>
>> to say HDCAM is 3:1:1, as 4:1:1 is perhaps more accurate in its
>
>> practical application. Same with the Varicam, it is 4:2:2 color space
>
>> in respect to 960 x 720. This leaves us with the somewhat awkward
>
>> ratio of 4:1.5:1.5, which is the color space in respect to the
>
>> originally captured 1280 x 720, and also the 1280 x 720 produced after
>
>> decompression of the signal and upsampling.
>
>>
>
>>
>
>>
>
>>
>
>>
>
>> Victor Wolansky
>
>>
>
>> DS VFX Artist
>
>>
>
>> WEBsite <http://www.e3post.com/>
>
>>
>
>> Demo Reel <http://www.e3post.com/reels/victorreel.html>
>
>>
>
>> 815 Slaters Lane
>
>>
>
>> Alexandria. VA.
>
>>
>
>> 443-797-3507
>
>>
>
>>
>
>>
>
>>
>
>>
>
> ---
>
> Subscribe? E-mail Majordomo(at)Softimage.COM with the following text in body:
>
> subscribe ds
>
> Unsubscribe? E-Mail Majordomo(at)Softimage.COM with the following text in
> body:
>
> unsubscribe ds
>
> ---
>
> Subscribe? E-mail Majordomo(at)Softimage.COM with the following text in
> body: subscribe ds
>
> Unsubscribe? E-Mail Majordomo(at)Softimage.COM with the following text in
> body: unsubscribe ds
>
---
Subscribe? E-mail Majordomo(at)Softimage.COM with the following text in body: subscribe ds
Unsubscribe? E-Mail Majordomo(at)Softimage.COM with the following text in body: unsubscribe ds
<<winmail.dat>>
