Περι graphics card....(4:4:4)

Re: Περι graphics card....

Xρήστο νομίζω οτι το subsampling είναι είδος διαδικασίας συμπίεσης σήματος βίντεο και δεν έχει σχέση με κάρτα εξόδου σήματος βίντεο. Απο ότι νιώθω δεν αφορά τις καρτούλες μας αλλα τα μηχανάκια καταγραφής-συμπίεσης-αποθήκευσης βίντεο. Λέω τώρα, μην το δέσεις.
 
Re: Περι graphics card....

Έχω ζοριστεί λίγο στο ψάξιμο και ας είμαι καλός "Browser" :grinning-smiley-043

Για δες αυτά και συνεχίζω το ψάξιμο...

http://www.top-shareware.net/Etymonix_MPEG2_Video_Codec.html
http://72.41.60.12/en/support/sheervideo/user_manual/sheer_codecs/sheer_ycbcra_10bv_4444.php
http://www.opencores.org/forums/video/2002/11/00003
http://forum.doom9.org/archive/index.php/t-69866.html

Υ.Γ. Αν έχω χρόνο θα ψάξω και λίγο το avisynth...
 
Last edited:
Re: Περι graphics card....

Απ' ότι βλέπω πάντως δεν υπάρχει καρτούλα που να υποστηρίζει τέτοια έξοδο... Μόνο με απ σέιμπλ γίνεται... (Δεν ειμαι σίγουρος ακόμα..)
 
Απάντηση: Περι graphics card....

Καταρχην, τοσο τα DVD οσο και τα HD-DVD/BluRay ειναι mastered σε 4:2:0, οπότε αν αναπαραχθουν απο ενα player με δυνατοτητα εξοδου 4:4:4, το υψηλότερο color sampling ουσιαστικά ειναι εξτρα πληροφορία που δεν προέρχεται απο τον δίσκο . Εχω ομως καπου διαβασει οτι η εξοδος σηματος 4:4:4 σε ενα προβολεα που να το υποστηριζει, ενδεχεται να εχει ευεργετικα αποτελεσματα σε θεματα color banding.
Αναφορικα με το θεμα αν η καρτα γραφικων μπορει να υποστηριζει εξοδο σηματος 4:4:4, υποθετω πως ναι, αφου υποστηριζει εξοδο RGB το οποιο ειναι εκ γενετης ασυμπιεστο 4:4:4.
 
Οποιος εχει ορεξη για διαβασμα, υπαρχει ενα καταπληκτικο σχετικο thread στο Avsforum, που αποδεικνυει γιατι θεωρειται αυτο το φορουμ "θησαυρος γνωσεων":http://www.avsforum.com/avs-vb/showthread.php?t=720502

Παραθετω μερικα απο τα πιο ενδιαφεροντα ποστς:

Maybe a bit of background here would be useful.

Today, we "subsample" the resolution of "chroma" (color) information in video to gain some efficiency. Non-subsampled video would be 4:4:4. However, the eye is not as sensitive to high frequency information in color domain(chroma), versus black-and-white ("Luma"). For example, you may see the difference between two shades of gray but not two shades of red. By subsampling, we essentially smear the edges of sharper color transitions. But if you eye couldn’t have seen that edge to begin with, you are not missing much.

Before you get worried about loss of resolution
smile.gif
, let me say the standard in broadcast world is 4:2:2. And this includes post production and transmission within the facility. The main application for 4:4:4, is to chroma-keying. Here, we are trying to isolate a color and replace it with another image (think the weather map behind the reporter). Needless to say, we want to find the edge as accurately as we possibly can, between say a blue screen and the talent’s cloths. Smearing the color by subsampling would make this job harder, causing the “key” to either bleed into the subject (parts of the hair missing) or not close enough (blue halo around the hair). So one prefers to work in 4:4:4 but even here, a lot of such work gets done in 4:2:2 anyway.

So I would say 4:2:2 should be the desirable next step if you don’t want to go crazy.

On 8-bit vs 10-bit, one advantage is workflow improvement as our source is 10-bits often so why convert if we don’t have to. Second reason is that good conversion relies on “dynamic rounding” or some other technique to better blend the transition as to avoid banding. Such smoothing at the edge may be more difficult to encode, and cause additional compression artifacts in that segment.

So I think 10-bit is a good goal to have. Putting it all together, I think the next step up should be 10-bit, 4:2:2. And maybe make 4:4:4 optional while we are at it.

Now, would most people appreciate this step up? That is the good question when folks claim upsmapling DVDs are almost as good
smile.gif
.

__________________
Amir
Microsoft (HD DVD insider)


-----------------------------------------------------------
A few comments:

1) 4:2:2 by itself is not enough. At HD resolution, it's very difficult to tell the difference between 4:2:0 and 4:2:2 on natural (not test pattern) content.

2) 10 bit would be nice, but it's a real chicken and egg problem. No encoder development will happen without a decoder. No decoder company will build a 10 bit decoder unless they see a market for it. Decoder margins are often pretty low, so unless the decoder manufacturer thinks that they can sell a boatload of chips, it's a tough sell to even consider the development effort.

3) The 4:4:4 profile in H.264 has been withdrawn.

Ron

__________________
HD MPEG-2 Test Patterns http://www.w6rz.net

------------------------------------------------------

4:2:2 and 4:2:0 exist to save data bandwidth, trying to take advantage of the way the human eye sees. Very little in production is preserved to the 4:4:4 level to the end. The only HD decks I'm aware of that support this kind of data rate are HDCAM SR decks. At my company we do a lot of HD production and most of it is on 4:2:2 D-5 decks. For a delivery format, I can completely understand the choice of NOT including 4:4:4 support. I think differences in the ammount of compression are going to be vastly more noticable than the reduced color bandwidth.

See the page on wikipedia on Chroma Subsampling for a nice explanation.

Since 90+% of everything being produced is 4:2:2, what benefeit would there be for 4:4:4 support in the next generation of HD DVD formats? How many people are going to have a screen and a signal chain clean enough to even notice a difference? You'd be hard pressed to see the difference when playing back clips on our $40,000 Sony HD monitors. You would notice the difference when trying to pull a key of a greenscreen background, but that's not something you're going to see at home on your TV anyway.

You can argue all you want about the data rate not being higher, but wanting 10-bit 4:4:4 is just silly. That's all I'm saying.

---
A full 10-bit clean path is not often maintained all the way through the end. You would have to have an extraordinary eye to tell the difference in motion video. You may actually see some banding in some certain color gradients, but this is rare, and subtle gradients is pretty much the only place where it's going to be noticable. The real benefeit for 10-bit video comes for those of us doing things to the image such as in visual effects. As a delivery format, on a monitor, 99% of the time you'd never know the difference.

KDogg765
-----------------------------------------------------------------
Hardly any point moving from 8bit to 10bit with a video intensity range. The intensity representation available in a video colourspace pretty much fits smoothly into 8bit without popping above the mean perceptability of your average human bean. Most consumer screens I've seen can't even handle an 8bit source perfectly anyway let alone 10bit. And the ones that could were all CRT I'm afraid.

More of a case for improving the colour sampling but to be honest it is a very efficient , minimally impactful way of saving on bandwidth especially with the move to natively progressive formats.

I seriously doubt on even high end consumer kit you will see any percieveable difference with 10bit vs 8bit and if you did it would probably only affect a tiny percentage of the images you would be viewing. I suspect you'd struggle to see much difference on pro level kit either. You might observe a difference with 4:4:4 or 4:2:2 vs 4:2:0 but again it would be dependant on image content and hardly a blanket improvement.

I would have prefered us to move to an 8bit log encoded 4:4:4 format that was more along the lines of a film type colourspace ( not necesserily full neg density like a 10bit log film scan but something nearer to print) fix all those flat video blacks and harsh whites we live with in video land.

I bet 1280x720 with that sort of improved colourspace would look significantly better than 1080p video.

The only trouble would be finding displays that could make the most of it . 10bit log certainly looks nice on the common 2k 3chip dlps and dilas but you hardly find many of them dotted around the place.

We need to dump the video intensity range and stop harping on about improvements to resolution and bti depth in my humble opinion.

__________________
digital film janitor

--------------------------------------------------------------
 
Top Bottom