[Demo] XviD HDTV ... ;-)

[DevilsChild]

The Terminator clip may not be the best source to test with. I re-encoded it to VP6 at 5092 kb/s with 2-pass Best Quality and saw the same "blockiness". Then I noticed that the source clip has quite a few blocks itself, mainly in the parts near the beginning and end with the blue flames.

Besides, I activated the VP6 "spatial resampling" function, which probably exaggerated the blockiness which was in the original clip. Overall it looks very good. Hard to believe XviD would have much trouble getting "decent" quality.

[bond]

i stumbled over this news entry:
http://www.theinquirer.net/?article=13406
hdtv mpeg-4 capturing with divx5 and xvid

software screenshots (http://www.countdownhd.com/Countdown_HD.htm)

[communist]

Originally posted by bond
i stumbled over this news entry:
http://www.theinquirer.net/?article=13406
hdtv mpeg-4 capturing with divx5 and xvid

software screenshots (http://www.countdownhd.com/Countdown_HD.htm)
Hehe...150$ and what exactly do I get for it?
'Encoding existing video..." isnt that possible with just about any other encoder like say VirtualDub?
Whats the point of that software? :confused:
All it does is transcode MPEG2 -> MPEG4 ?!

[mf]

And the screenshot is from a badly deinterlaced interlaced source. Bad publicity :rolleyes:.

[snoddas]

Sagittaire,

do you still have the XviD-HDTV-Q4.rar clip anywhere on the net? I can't find it using the original link you supplied.

[bob0r]

Originally posted by snoddas
Sagittaire,

do you still have the XviD-HDTV-Q4.rar clip anywhere on the net? I can't find it using the original link you supplied.

http://66.246.16.28/prosac_upload/tmp/XviD-HDTV.avi
Temporary online, grab it while you can...

[snoddas]

thanks bob0r!

[kadajawi]

wow, thanks. Awesome quality. Although it looks a bit blocky at times.

[unmei]

you guys could know this: what is the max picture dimension for xvid? (or max pixel in a frame)
...i thought it were 4096x4096 but recent problems make me think it's lower...

i know this were a classic for search, but it seems like my seach patterns fail on this ..either i get a million unrelated threads or none at all :(

[communist]

AFAIK:
MPEG2 allows for resolutions up to 16000x16000.
I wouldn expect MPEG4 (XviD) to be lower than that :)

/Edit looks like MPEG4 allows res to go up to 4096x4096.

[sxd]

I took a stab at this as well, using a DVD source (Harry Potter 2 trailer, same as in this thread. (http://forum.doom9.org/showthread.php?s=&threadid=64789)

-TomsMoComp(0, -1, 0) for resizing from 720x304 to 1440x806
-Constant Quantizers (2 for I/P, 4 for B)
-B frames 2/1.50/1.00, Closed GOV
-MPEG, QPel, Trellis, Chroma motion, MS 6, VHQ 4
-Bitrate about 8.7Mbps (about .3 bits/pixel)

Here's 10MB of it. (http://www.tc.umn.edu/~plim0005/hp2-hd.avi)

Overall I think it's ok, but would be nicer if I could play it back :rolleyes:

TMC is not the greatest resizer, but I had to give it a shot.

edit: Reencoded, exhibits same problems... probably just too much for my machine.

[communist]

Odd... that original HDTV demo ran pretty fine here - not the same with your Harry Potter sample.
Oh and how do you rsize with TMC?!:confused:
I thought it was a deinterlace.

[sxd]

Well, it is -

SearchEffort - determines how much effort (CPU time) will be used to find
moved pixels. Currently numbers from -1 to 30 with 0 being practically
just a smarter bob and 30 being fairly CPU intensive.

For Avisynth only, a value of -1 is supported. In this case the TomsMoComp
filter will not deinterlace but instead assume you already have progressive
frames but want to double the vertical size. I found by accident that this could
give slightly better apparent detail than regular scaling algorithms and is useful
for low bit rate captures that are hard to IVTC/deinterlace or where you have just
kept the even fields for some other reason. I'm considering making a DirectShow
version of this to be run at display time, or possibly adding it to ffDshow.


explains it, straight from the readme.

I'm not really sure why it's so choppy, maybe packed bitstream would help. There is a large fraction of B-frames, so that might be it too.

Well, part of why it's choppy is there are actually frames missing (???). When the bird hits the bowl of chips, it jumps like 10 frames ahead to where the bird is on its back, instead of showing the whole fall. I used VirtualDubMod to chop 10MB out of the full trailer, and it seems it's royally screwed up. When I seek to keyframes, the keyframes don't have the right number and occasionally appear out of order despite being able to watch it in MPC without a problem (aside from 10 fps).

[kilg0r3]

Well, their was some discussion about this question in relation to 'Xvid-HDTV-Encoding Tests'. So I thought I'd give my two cents here too.

IMO, the advantage of encoding dvd material square pixel resolutions of about 1080 is due to the fact that any recompression (xvid) will introduce artifacts; e.g. ringing.

If you enlarge the video only during playback, you will also enlarge the artifacts introduced by the recompression; most notably vertikal edge artifacts (or double edges). If you enlarge before recompression only those artifacts that are already in the source get enlarged.

Of course, an effect of increasing the resolution will be that their will be an increase either in bitrate consumption or in the number of artifacts introduced by recompression. However, due to the fact that you do not add new information just by resitzing the source, it _might be that_ the increase be considerably lower than expected. This is the first time I come to think of this fact in a positive way.:) Furthermore, the higher the resolution the smaller the blocks.

What do think of this?

[bot]

Wouldn't it be easier and perhaps more shrewd to /not/ re-encode the DVD, but enlarge it during playback?

[kilg0r3]

but then i would not have a backup of my dvd ....

[SeeMoreDigital]

I did not know this thread existed.

It really is quite pointless generating 'so called' HD encodes when they were created from none HD sources, such as an DVD.

As some of you may be aware I've generated some XviD tests files of my own from genuine HD sources ie: 1080i Mpeg2/AC3 and AAC. And even started a thread about it in the 'General' section of the forum.

I concluded that after deinterlacing an 1080i source there really was no need to keep the vertical pixel height at 1080 high so I resized it to 720 high (as used for 720p broadcasts). As not only will the average PC struggle to play an 1080 XviD encode (occupying 1920x1080 = 2,073,600 total pixels), they don't actually look as good as the 720 high (1280x720 = 921,600 total pixels) resized encode anyway. And I guess there are other technical reasons why when an 1080i source is de-interlaced to 1080(p) it does'nt look so hot!

Now, I did not get around to posting for download any of my 1280x720 XviD encodes because I elected to generate and post some anamorphic versions instead using just 576x720 pixels (within an MP4 container).

Many people reported that they looked pretty darn good, which is quite surprising when you consider how far the 576 horizontal pixels are being stretched. I originally only generated them at this weird size so I could play them in hardware (via Sigma Xcard) and view the results on a big TV screen.

Sufficed to say the 1280x720 'True 16:9 Framed' XviD encodes I generated look better!

One more thing. I can't help smiling when I read suggestions that some people want to resize their HD captures to image pixel frame sizes they are currently using for their DVD backups.

The whole point of HDTV is viewing the images on a big screen. Not being stuck behind a PC monitor. OK I accept that there will be some people who will want to re-encode an 2.35:1 image (via a 1920x1080i HD source) down to 640x272.... but please don't try and pass this off as being anything like as good as the source, it's just not worth debating!

Cheers

[mf]

Originally posted by kilg0r3
IMO, the advantage of encoding dvd material square pixel resolutions of about 1080 is due to the fact that any recompression (xvid) will introduce artifacts; e.g. ringing.
What do think of this?
Yeah, I observed and noted this on my 1600x864 reloaded trailer. In standard encoding terms, it would be rated pretty low quality. It has a lot of ringing and in high motion the image just breaks apart in blocks. However, two things happen. The blocks are only 8x8 on a 1600x1200 screen, so you don't even notice them. And if you did, the postprocessing works so good that the deblocking looks virtually perfect. I found that pretty interesting :).

[SeeMoreDigital]

A lot of the effects you are seeing are more likely caused during de-interlacing (that is if you are encoding from an 1080i source).

Not forgetting that de-interlacing from NTSC sources is a more complex procedure than from an PAL source.

Cheers

[bot]

Originally posted by kilg0r3
but then i would not have a backup of my dvd ....


But as SeeMoreDigital hints at, you can't introduce new information by enlarging the image before transcoding it.
-If you strip and rip your dvd to your hard drive, perhaps even as one big .vob and renaming it .mpg, the image quality would be maximized.

I haven't tried encoding HiDef xvid myself (yet). Neither from HDTV broadcast, nor from upscaling DVD's.
-If you compare the file sizes of stripped DVD's and upsampled xvid transcodings, what kind of difference do you see (percentage wise)?

[SeeMoreDigital]

Ok I've taken the liberty of generating and posting some XviD encodes I've created from an genuine 'interlaced' NTSC 1920x1080 HDTV captured source - None of your up-sized 'progressive' NTSC 720x480 DVD rubbish here!

Both have been encoded using 2 passes at around 3000kbps. Meaning that approximately 150 mins could be 'backed up' onto an 4.36GB DVD~R/RW.

This 16.9MB (http://homepage.ntlworld.com/seemoredigital/MichelleBranch_1280x720.zip) encode was generated with an 1280x720 (true 16:9 frame) image pixel frame size.

This 17.6MB (http://homepage.ntlworld.com/seemoredigital/MichelleBranch_576x720.zip) encode was generated with an 576x720 (anamorphic frame) image pixel frame size.
If you have a widescreen TV and an Sigma Xcard, you'll like this encode, because it will automatically open up at the correct aspect ratio.

As you can see 'both' use the same quantity of vertical pixels.

Please feel free to judge as you see fit!

Cheers

[kilg0r3]

O.k. guys,

first of all, I don't know a thing about the HDTV format. Thus it does not take wonder that I, again, misunderstood some pretty basic things, e.g, I thought that 1080i designates a frame which is 1080 pixels _wide_ :blush:.

What I was actually talking about is the horizontal stretching of a 720(h)*576(v) source, which, after cropping of the black borders, might give you a frame size of 896*496 for a 16:9 movie. This, as you will have already grasped, does not actually have anything to do with HDTV encoding. So, you can bash or bann me now for blathering about nonsense in the wrong place.-If you compare the file sizes of stripped DVD's and upsampled xvid transcodings, what kind of difference do you see (percentage wise)?
I havn't done extensive testing, but I am pretty sure that you can get away with 50% or less of the filesize.

[Chainmax]

SeeMoreDigital, I can't download any of your clips.

[SeeMoreDigital]

Originally posted by Chainmax
SeeMoreDigital, I can't download any of your clips. Yep, sorry about that, I'm currently uploading them to an faster ftp server instead of having them stuck on my home server!

They will be up soon. Probably around 19:00 UK time.

Cheers

Everything should be working ok now

[Chainmax]

Yup, downloads work now.

[bob0r]

Those .mp4 files, what codec is used for that?

I mean, windows media player tries to download some extra codec?

[bond]

Originally posted by bob0r
Those .mp4 files, what codec is used for that?

I mean, windows media player tries to download some extra codec?if yould have used search you would have found this (http://forum.doom9.org/showthread.php?s=&threadid=62723) ;)

[SeeMoreDigital]

Originally posted by bob0r
Those .mp4 files, what codec is used for that?

I mean, windows media player tries to download some extra codec? XviD for the video and AAC for the audio!

[Joe Fenton]

... and 3ivX for the mp4 splitter.

[SeeMoreDigital]

Originally posted by Joe Fenton
... and 3ivX for the mp4 splitter. Yep, there are quite a few solutions for getting .mp4 contained audio and video to work.

I still favour installing the promo version of Nero's Vision Express, as you also get a trial period of Recode2 along with all of Nero's filters, splitters etc.

That said, Nero's installation is not perfect as it still can't handle some 'anamorphicly encoded' content correctly. But I have created a fix for this that you can download here (http://82.2.167.24/Uploaded_Files/Doom9_Forum_files/Nero_DSFilter_Fix.zip). Just follow the enclosed instructions.

Once installed all .mp4 content will play ok via Windows Media Player 9, even anamorphicly encoded content.

Cheers