Does transcoding lower the bitrate?

[timmyp]

As the subject says, does transcoding lower the bitrate? Or does the file size reduction come from some other factor?

Thanks!

[adam]

http://forum.doom9.org/showthread.php?s=&threadid=63587

This is a very good explanation of how transcoding works. Information is thrown away so yes, the bitrate is lowered.

[Redman66]

with this one exception...

[Redman66]

well the picture wouldn't load , so here's the link that describes transcoding without a bitrate change....

http://www.bbc.co.uk/rd/pubs/papers/paper_04/paper_04.html

[adam]

Well no you don't have to lower the bitrate, but then you don't gain any compression either.

[timmyp]

OK, thanks to both of you for the links.

A few more details.... I have some mpg's that apparently were coded at 9800 MB/s and I'm receiving errors in TMPGEnc DVD Author that these may be out of compliance for a DVD (audio is full 1536 PCM, bringing total bitrate over 11,000).

Would it be better to use DVD Shrink/DVD2ONE to transcode these file after creation with TMPGEnc DVD Author? Or recode them with TMPGEnc before authoring? (Need to keep audio at PCM due to music in mpg's.)

Moderator, please move this thread if needed.

Thanks again.

[Cornucopia]

Well, the difference between so-called "transcoders" and re-encoding can be somewhat subjective, but supposedly, a transcoder would have the edge over the re-encoding because it doesn't need to completely decode and uncompress the program, it catches it mid decompression and just changes the DCT quantizing coefficients to allow for (usually) greater compression. This is good because it retains all the other qualities of the 1st compression pass and thus, minimizes the compounding of the "error".
But as usual, YMMV.

Note: how important is the audio vs. the video. You could always convert the audio from 1536PCM to say, 384 or 448 AC3 and save a good chunk for video. Then it wouldn't have to be lowered as much (from 9800 to ~9200). At that high rate, even re-encoding wouldn't be very noticeable.

Scott

[adam]

Sorry but there is no way a transcoder has an edge over an encoder in regards to the "error" or accuracy of the resulting stream. It is the exact opposite. The only thing a transcoder gets to keep is the motion vectors. It is the quantization which is so essential because this is what results in the actual artifacts that people see. By re-quantizing the co-efficients of the bitstream it loses substantial accuracy as compared to a re-encoding. Transcoding is a shortcut. It avoids the lengthy process of computing motion vectors at the expense of accuracy and quality of the bitstream.

Assuming anything more than miminal compression, requantizing co-efficients as opposed to re-encoding the stream will ALWAYS result in a less accurate bitstream and thus a lower quality stream. Whether or not you notice the quality difference is subjective, but the fact that re-encoding produces a more accurate representation of the source stream is a fact.

Another fact to consider is that re-encoding allows you to divide quality loss, more or less, equally among all frames. If you wish to compress something by 40% you generally wouldn't want to only compress say, the last half of the movie. But with transcoding the only thing that can be compressed is DCT data, and each picture has a different ratio of DCT and motion data. So there will be many frames in the stream which either cannot be compressed much or at all. As a result, some frames will be left wholly uncompressed, while others will have to be compressed double to compensate. This is why transcoders result in very specific artifacts. You may get a flawless scene which is essentially unchanged from the source stream, and then out of nowhere get a huge block error dead center in the screen.

[Cornucopia]

adam-
sounds reasonable (esp. the last part), but do you have any documentation or links to back that up? it'd make me feel more solid on the whole thing.

thanks,
Scott

[bugster]

OK, thanks to both of you for the links.

A few more details.... I have some mpg's that apparently were coded at 9800 MB/s and I'm receiving errors in TMPGEnc DVD Author that these may be out of compliance for a DVD (audio is full 1536 PCM, bringing total bitrate over 11,000).

Would it be better to use DVD Shrink/DVD2ONE to transcode these file after creation with TMPGEnc DVD Author? Or recode them with TMPGEnc before authoring? (Need to keep audio at PCM due to music in mpg's.)

Moderator, please move this thread if needed.

Thanks again.

There have been sveral similiar threads to this recently regarding Tmpgenc DVD Author and max bitrate. It appears that TDVDA only looks in the mpg (or vob) header for the nominal bitrate. Often this is listed as 9.8Mbps, as this is the DVD max, even when the actual bitrate in the mpg is much lower. I suggest using bitrate viewer on your source file. If the bitrate does not exceed 8Mps, you are safe to use the LPCM audio and ignore the TDVDA warning.

[adam]

Cornucopia, well I thought most of it was explained in the link I posted. Granted its just a forum post but the information appears very accurate, especially since the latter half of it is a direct quote from the author of DVDShrink. Nevertheless, I'll try to find some technical documents which lay it out a little more thoroughly.

Re-encoding quantizes by analyzing each and every frame in the source and throwing out the most redundant data. Transcoding requantizes almost completely randomly. Obviously the latter is going to be less accurate.

Honestly the proof is in the pudding. There is no question that re-encoding produces better quality at higher compression levels than transcoding, because its just been tried and tested too many thousands of times to be a fluke. Try comparing the results of any one of the popular transcoders compared to a re-encode of any one of the popular encoders. At low compression ratios the difference is going to be transparant. But as that level increases the difference is going to become undeniable. Try comparing a DVD transcode at 50% to a re-encode at 50%. The transcode will look horrible and the re-encode just may look visually almost identical to the source. Hell even the author of DVDShrink says that, "DVD Shrink sucks at 50% compression." Yet I frequently compress DVDs by 50% using CCE and I honestly see no difference in the result.

[spectroelectro]

I would strongly agree with the previous poster except for one program that does an exceptional job of transcoding: ReJib. I've crunched MPEG-2 files by to as little as 60% of the original size with no visible artifacts, and those were pretty high bitrate files. Best of all ReJig is free. It's a vastly superior product to DVDShrink, by the way. If you need to really slam the size down, then re-encode. But you can get all the way down to about 60% or so of the original filesize with ReJig and I can't see any difference twixt the original and the ReJigged file.

[timmyp]

Bugster: Thanks for the info on TMPGEnc DVD Author... that got me heading in the right direction to look at the ACTUAL bitrate of these mpg's. I ended up doing a re-encode with TMPGEnc down to a bitrate of 6000 (looks fine to me!) and was able to fit them onto one DVDR with the PCM audio.

Thanks to all who weighed in and helped steer me towards a re-encode rather than transcoding.

[adam]

I agree that Rejig has the best transcoding quality out of all the other programs, but its still inherantly inferior to a re-encoder. I personally can see a difference when the compression level gets high enough. Also, the higher the source's bitrate, the LESS noticable any further compression will be. Compressing a very high bitrate source by 60% may be relative to compressing a typical source by only 5-10%. You really need to look more at the average bitrate that you are ending up at. If your output is in the 3-4mbit range and you still don't see a difference than that is good. Above this and there really shouldn't be a difference.