Stepping on MPEG2 - how bad a loss?

BJ_M wrote on 11/15/2004, 3:39 PM

depends on the quality of the source to begin with -- you would be the best judge of your finished product because we can't see the before and after..

but I have rendered out plenty of mpeg2 files and they can look very very good --- but never the same as the original (sometimes though can be better - if source is full of noise or bad colors or something) ..

the quality of the source material is the prime , most importaint factor in mpeg encoding , more than what settings or encoder is used.

Liam_Vegas wrote on 11/15/2004, 4:30 PM

Not only are you lazy... but you will be losing quality every time you re-render the mpeg. That's pretty bad in my book.

Also.. not quite sure what you are doing when you "fix the DVD it's on"? Could you explain that a little more please?

My quesion would be why is DVDA re-rendering that MPEG? I render my MPEG files using Vegas - load them into DVDA - and it never has to re-render.

I think the only reason it does re-render is if your project settings in DVDA do not match the settings for you MPEG from Vegas - or if you have chosen "fit to disk" in DVDA.

Jay Gladwell wrote on 11/15/2004, 5:22 PM

It was stated in another post that MPEG was compressed 25:1. So if you use that for your edit and recompress it another 25:1, it will be like compressing the original 50:1, at least, if not worse. It'll probably look pretty shabby.

Jay

BJ_M wrote on 11/15/2004, 6:02 PM

that isnt quite true - otherwise if you re-render lets say a DV stream 3 times you would be saying it is compressed 15:1 ...

before it is recompressed, it un-compressed first ...

Jay Gladwell wrote on 11/16/2004, 10:50 AM

BJ_M, that's not accurate.

Doing a "render" is very different from doing a "compression". Once a file is compressed, you're never going to regain what was lost in the compression. Hence, if you compress a file 25:1 and, after editing it, you recompress (not render) it to MPEG, you will lose again and it will probably look much worse because you have 25 times less material for the compression to work from than in the original .avi file.

Jay

BJ_M wrote on 11/16/2004, 12:24 PM

that is not true --

for example i do a compression of cin files to LZW targa files .. what you are saying is that I am throwing out 1/6 of my material ... no ..

say you compress a word doc with zip -- you dont lose anything when you unzip it ..

now for mpeg compression, a very lossy compression, except at higher bit rates , if you re-render it or re-compress it to mpeg once again , the render engine will have decode it first before it can be re-compressed, sure info was lost the first time it was compressed, but not automatically 25 times LESS ... it would depend how much you lost on a number of factors .. many in fact..

If you re-edit a mpeg and compress it out to mpeg again , it internally is rendered first...

if i lost 25 times less material everytime i have to refix a mpeg file , i would end up with nothing .. 25:1 is a arbitrary amount anyway ..

what you are saying would make IMX and HDV both crap formats , and most people think they are not ..

Jay Gladwell wrote on 11/16/2004, 12:31 PM

Neither is what you're saying.

We're not talking about LZW targa files or IMX files or HDV files. We're not talking zipping text files. We're talking about recompressing plain ol' mpeg video files into plain ol' mpeg video files. It simply ain't gonna work.

There may not be a 25 times reduction each time, but editing and recompressing mpeg files is not the way to go if you want to maintain image quality.

Jay

BJ_M wrote on 11/16/2004, 1:35 PM

but HDV and IMX ARE mpeg files, HDV is not a lot different than say a DVD if you factor in resolution vs. bit rate ....

sure you do lose quality when you re-encode a mpeg .... but it is done all the time and in many cases,,,, you can not tell the difference except by the most minute examination (depending again on a number of factors - big one being the source to begin with) ..

it DOES work --- believe me ... 100's of 1000's do it all the time when they back up dvd9 to dvd5 disks for example .. we have 1000's of people devoted to how to and how to increase quality to the utmost on our websites and forums (maybe not here) ..

i have to re-encode mpegs probably nearly every day .... they look very very close , and in some cases better (as i have to fix things like noise, color, levels, etc) than the originals .. I'm not the only one doing it believe me ..

sure -- it is not my preferred choice of source format and it is not meant to be an editing format (but it is) ..

farss wrote on 11/16/2004, 1:48 PM

BJ_M,
as I know you know a lot more about mpeg-2 than me maybe you can tell me if I have this right.
Assuming the encoder can encode without any residual then the encode is lossless. However if the encoder cannot then something gets lost.
Now I'm only guessing here but I'd assume if you start to get serious errors in the encoding then the next time it's re-encoder the encoder may face another set of unencodable frames and things get even worse.
This kind of goes back to what you've been saying all along, if it was well shot to start with then the first encode is going to be pretty close to lossless so the degradation as you go down generations will be minimal. If it was poorly shot (and without consider how mpeg-2 encoder works) then it fall apart very quickly.
Whener my clients say they want it for DVD I try to educate them about how to shoot to get a good result. One thing I've noticed mpeg-2 doesn't like at all is noisy blacks, a frame with mostly noisy blacks seems to bring out the worst in mpeg-2 encoding. Either not letting the camera get into that situation or cleaning it up prior to encoding makes a big difference.
One other thing not mentioned, you don't have to encode at 6 Mb/sec, there's no reason to not encode at say 19Mb/sec and burn that onto a DVD-ROM for archiving, just don't expect the average player to cope.
Bob.

BJ_M wrote on 11/16/2004, 3:38 PM

mpeg2 at the low bitrates used for dvd's will fall apart quickly as you have said if the source is "mpeg unfriendly" , this means levels, noise, smoke/fire/fog type of effects/lighting (or waterfalls!) , shaking (i.e handheld) , really fast zooms or pans .... a lot of care will be required to handle those scenes (or films) if they are bad enough.....

One of the reasons I'm not really a huge user of DV in fact, is that I don't find it high enough quality as a source file without a lot of work to to blow it up or re-re-encode it as mpeg over agian .. Can be done - but not work the effort and requires more work it seems to shoot it properly .. I have used DV though, and used dv footage transcribed to HD and film resolutions ..

to re-encode mpeg , the mpeg encoder itself becomes more critical also ..

A big (huge) trend born of the digital age of broadcast is to not re-encode mpeg streams but to transcode them, that does not involve (re)encoding at all but basically chopping off bitrate (a very good explanation of this process is on the web - it is quite technical how this works) ..
The advantage to this is that is VERY fast and broadcasters can stick more streams in their bandwidth at the same cost... disadvantages is that it often looks like crap and not as good as lengthy time consuming re-encoding ... Consumer DVD backup programs like DVD shrink and dvd2one use the exact same process to achieve their speed (take off 30% of the bite rate of a 2 hour movie in about 20 min.) .

Archiving at high bit rate is a very common thing ... HDV is normal 15frame GOP at 18-25meg/s and IMX is IPPP mpeg2 .... I would say that a good mjpeg type codec may be a better choice if quality vs. space is an issue though . But mpeg can be used -- a higher than 19 is good.. like 40-50 for 1080i (though you cut out B frames nearly at that point) ..
Still of course, archiving with a lossless format is better yet .....