Re-encoding and not losing quality?

Hello, my e-friends!

I have a question. I keep hearing and hearing (actually reading) that when we re-encode a video, we lose quality in the process. However, doesn't this only applies when we re-encode with different specifications? Let's say, without considering the audio, if we re-encode an MPEG-2 video, 720 X 480 size, 29.97 fames per second, and 8000 kbps constant bitrate to another MPEG-2 with exactly the same specifications, do we still lose quality in the process? I tend to think that losing ony happens when we re-encode to different specifications or to different video formats like from MPEG-2 to WMV, for instance.

If I am wrong, why does this degradation take place then?

CAL "Disco Mak"

Whenever you re-encode with a lossy codec or change specifications, you lose quality. Whether the quality loss is noticable or not is a different matter. So MPEG-2 to MPEG-2, you lose quality. MPEG-2 to wmv, maybe not since wmv supports lossless audio and video. Re-encoding lossy to lossless doesn't make a lot of sense though since the filesize increases and the quality stays the same. Can be usefull for instance if you have some heavy filtering and are doing multiple passes.

How about if you take an mpeg2 at 3700kbps and re-encode at 7000kbps? And what kind of data loss are we talking about? Is it color info? Motion (noticeable in fast-moving scenes). Detail? What are we going to notice first?


Darryl

Originally Posted by dphirschler

How about if you take an mpeg2 at 3700kbps and re-encode at 7000kbps?

The image quality will still degrade a bit.

Originally Posted by dphirschler

And what kind of data loss are we talking about? Is it color info? Motion (noticeable in fast-moving scenes). Detail? What are we going to notice first?

All of those things happen. The most obvious problems are macroblocks and ringing (or mosquito noise) around high contrast edges.

8x, nearest neighbor enlargement of mosquito noise:



This is an artifact of DCT compression and happens regardless of the bitrate chosen.

What people may forget is that to re-encode = encode + decode + encode. That means, 1st encode loses some stuff, decode doesn't lose anything--but it does play the lesser stuff AS IF it were always uncompressed AT THAT LEVEL.
And so, encoding the 2nd time, it sees an UNCOMRPESSED (but of lesser quality) image TO START WITH, and then takes out what it needs.

The good thing is, that image has less "stuff in need of taking out", so the 2nd and subsequent compressions are a little more efficient than the 1st--unless you're REALLY taking out alot (much lower bitrate).

Example:

Starting with good image = 100%

1st compression is bitrate of 6000kbps = 85%
De-compressed image is still 85%

Starting with that,
2nd compression is bitrate of 7500kbps (thinking "let's make this better) = 90%
Well, 90% of 85% is 76.5% -- STILL LESS.
If you had done 6000kbps again, it might have been 87% (because of less to take out),
but that still is 87% of 85% = 73.95%
Even 9800kbps isn't = 100%, it might be 98.5%
So 98.5% of 85% = 83.725%.

How to fix this?

Go back to the source and do a new 1st encode.

Scott

The other thing is that an encoder can't really tell the difference between artifacts and details, so it has to waste bits preserving artifacts and if you don't crop/resize to move the macroblocks, then you can compound the macroblocks.

Hello Scott and rest!

Thanks to each of you for your explanations.

Okay, Scott, I kind of understand what you are saying, but I have two questions:

1) Let's say that I have an MPEG file recorded from an analog VHS source. The original digital file is obviously the MPEG file but the original file is still the analog VHS source. If I work from the original digital file, will I be encoding or re-encoding? Needless to say is the fact that I cannot work from the original file, the analog file, since it is analog.

2) If we always lose quality no matter what, then, when we author DVD-ready files such as VOB, BUP, etc. from an MPEG source, will we also be losing quality?

Any responses are appreciated.

CAL "Disco Mak"
discomakberto@megamixers.co.uk

There are MPEG editors that will not reencode if you are performing simple cut and paste operations -- except for small segments right at the cuts which need to be reencoded. Or, if you add a transition between two clips, they will only reencode the frames during the transition.

If you perform any filtering (resizing, sharpening, color adjustments, etc) the filtered segment will have to be reencoded.

In short, anything that modifies the picture necessitates that the MPEG data be decompressed, the picture modified, then the picture be recompressed.

Converting to VOB to MPEG is just a different multiplexing, or layout, of the audio, video, and other other data. The MPEG data isn't decoded for this, just "repackaged".

I concur with jagabo, with this addition:

Authoring may include encoding (if it's necessary to make compliant) or just be re-organizing (if it's already compliant).

And, Yes, VHS is analog and so isn't a "file" such that it can be reworked. Nonetheless, if your 1st generation from VHS is directly an MPEG, and it isn't exactly what you want in terms of specifications, then IIWY, I'd re-digitize the VHS to the correct specifications.

(Of course, I almost ALWAYS digitize to uncompressed/lossless/MJPEG/DV AVI/QT first anyway).

Scott

I'm talking about Authoring programs/packages, not "authoring" the process.

Scott

I think it is possible to encode mpeg to mpeg without loosing quality as below:
use all P frames no I or B frames
quant matrix all 1's

Warning do not try this at home.
It may just fill up the rest of you hard drives free space

There is another way, go into explorer and cut and paste

Originally Posted by j4gg3rr

I think it is possible to encode mpeg to mpeg without loosing quality as below:
use all P frames no I or B frames
quant matrix all 1's

Warning do not try this at home.
It may just fill up the rest of you hard drives free space

There is another way, go into explorer and cut and paste

No I-frames???? That's not a valid GOP.
And QuantMatrix of 1's is counter to "saving space", which is what MPEG is all about.

Yes, let's "don't try this at home".

Scott

Sorry all I frames no P or B.
But does that loose quality at all, maybee in the search precision it still does?

Originally Posted by j4gg3rr

Sorry all I frames no P or B.
But does that loose quality at all, maybee in the search precision it still does?

All I frames still loses quality. I frames are essentially jpeg images.

Hello,

I have been following this discussion and I'm wondering :
- suppose you do "original -> codec -> decode -> re-encode" and
you use the same codec for 're-encoding' as you did for the
'original' compression and the same quality, will there be a
difference between the 2 coded files ?