Sample DVD for comparison purpose

[Makira]

Since we can't post screenshots of commercial DVDs, it can be difficult to compare various methods for recompressing DVDs. So I built a small DVD just for this purpose.

Anyone may redistribute it, it is free. (in fact, if someone wants to host it...)

Features 3 small clips, one pan of a mountain view, one shot of a fire, one shot of a baby (sister's daugther). High bitrate two pass vbr mpeg2, interlaced (camera doesn't support progressive, sorry), raw images unfiltered.

It's 68 mb.
http://www.info.polymtl.ca/~anmis/SampleDiscDVD.tar.bz2

[dragongodz]

thanks Makira i will download this and have a look. though of course captured footage is not exactly the same as dvd footage. then again there are also different quality dvds aswell. :)

[Nic]

dragongodz, maybe if you get a chance we could do tests using this clip for different ReJig engines, and see how they come out? Would be quite interesting for us all to be working off the same material...ill host the file as well.

-Nic

ps
mirror: http://nic.dnsalias.com/SampleDiscDVD.zip
(this doesn't have the AUDIO_TS dir in, so you may need to create an empty one for compatibilities sake)

[dragongodz]

Nic - yes that is what i was thinking aswell. :)

[dvdshrink]

I noticed the VOBS contain two or three padding packs. Strip those and you get a little compression for free!

A mixed DVD with interlaced and progressive material would be wonderful. DVD Shrink for one doesn't like interlaced material very much, also since most DVDs encode the main movie in progressive format, I haven't spent much time trying to improve this case.

The sample DVD is however a great combination of scenes.

[Makira]

The disc was mastered by DVD Studio Pro, which introduce these padding block (I do not know why). Removing them gives compression for free indeed, but I also discovered bugs with them (for encryption detection - should not expect the first pes to have an extended header..).

For progresive stuff, I'll try to get my hand on a progressive camera at university or elsewhere and add a few clips. Give me a few days.

[JuanC]

Originally posted by Makira
... Anyone may redistribute it, it is free. (in fact, if someone wants to host it...) Could we use some P2P tool? may be eMule?

[dragongodz]

"A mixed DVD with interlaced and progressive material would be wonderful"
i agree totally.

"since most DVDs encode the main movie in progressive format"
i will have a quick run though a bunch of my dvds tommorow(probably the better known ones like terminator etc) and see what percentage are progressive. most does not always mean a large difference. :)

[int 21h]

Maybe have a look here: http://www.archive.org/movies/movies.php

Most of them are available in Mpeg-2 format, and I'm sure somewhere in there you'll find a progressive clip or two.

Plus all these are in the public domain :)

[dragongodz]

ok ripped some first vobs of a random selection and run through dvd2avi and bitrate viewer. these all came up as reported to be interlaced.

big trouble in little china
terminator 2
highlander
rugrats in paris
red dwarf season 3

so while some main movies etc are progresive some are interlaced. thus any realistic tests needs to use both.

"DVD Shrink for one doesn't like interlaced material very much, also since most DVDs encode the main movie in progressive format, I haven't spent much time trying to improve this case."
hopefully this will encourage you to. :D

[dvdshrink]

OK, point taken!

I only have two interlaced DVDs out of about 50 (both concerts). The rest are NTSC frame encoded, telecine.

Technically DVD Shrink doesn't care if a source is interlaced or progressive - the former just means twice as many encoded pictures - but the artifacts from requantization do appear worse on interlaced material. Maybe this is due to the half-resolution of each picture? It is definitely worth investigating.

[dragongodz]

"the artifacts from requantization do appear worse on interlaced material. Maybe this is due to the half-resolution of each picture?"
i would imagine that if the requant of a section of a field was quant revalued a different amount to the other field this would show up badly when the fields were combined on playback. producing what would appear as artifacts.