MPEG-2 Transcoding/Shrinking

[BobaFett77]

Hi everybody,

I'm recording TV programs with a digital satellite receiver in order to archive them on DVD-Rs later on. I noticed that some of my files were too big for a single DVD-R, so I decided to reduce the size of some of them (instead of putting them on two DVD-Rs).

My question is: What are your experiences doing this?

I first re-encoded the m2v video stream of one recording with TMPGEnc, but noticed a clear loss in image detail. In addition to that it took a few hours.

I then tried ReJig, which transcoded the MPEG-2 video from 5 GB to 4 GB (~80%). This was much faster and at least with that recording (a theater play without many action scenes) the result was great, I couldn't really see a difference to the original. But I found out that DVDlab also has a transcoding option, a DCT transcoder that quickly compresses MPEG-2 video files.

I'm wondering which program has the best results in general. I also read about DVD Shrink, which can be used to reduce the size of DVD projects that are already compiled. I think I also read somewhere that InstantCopy has good results (better results than ReJig, is that true?). There are guides for many programs and many uses online, but I'm unsure which of them is the best for reducing the size of DVB records.

What are your experiences?

[mats.hogberg]

Generally, reencoding is better but slower that transcoding.
Among the transcoders I've tried (ReJig & DVDShrink) I've found no noticable difference in quality.

/Mats

[guns1inger]

As image quality truely is in the eye of the beholder, why not let DLP transocde a segment for you so you can compare for yourself. By all means ask what others are using, but never take anyone's word for the quality. There are people who will argue till you go blind that the artifacts were not in the original overcompressed downloaded avi. Of course they were, but some choose not to see them. Only you can make the final call on what works for you, and that takes time. And it will continue to take time because there will always be new tools and methods to try, and occassionally migrate to.

[jbodin]

If your satellite receiver is a TiVo or DISH Network PVR, then your files are probably 480x480 resolution -- I've had great luck re-encoding these type of files as 352x480 resolution with a bitrate of around 5000 using gui4ffmpeg. Using this method, one cartoon movie (1 hour, 16 minutes of programming) compressed down to a 1.1 GB file; another one-hour program (43 minutes of programming after editing out commercials) compressed down to a 685 MB file.

If the native resolution of your satellite images is only 480x480 to begin with, you'll notice virtually no image degradation by going to 352x480, and you'll be able to fit more shows on a single disk.

[mats.hogberg]

...or, your player may play them anyway, if you run them thru svcd2dvdmpg + author (or better yet; SVCD2DVD)

/Mats

[ronnylov]

...or, your player may play them anyway, if you run them thru svcd2dvdmpg + author (or better yet; SVCD2DVD)

/Mats

...or DVDLab which has built in support for putting SVCD on DVD.

[mats.hogberg]

...or DVDLab which has built in support for putting SVCD on DVD.

Rather - doesn't enforce standards.

/Mats

[Tom Saurus]

BobaFett77: I suggest you go to Lordsmurf's website and follow some of his guides there to get good results recording using your capture card. I have a ATI All In Wonder 9800 Pro and I followed his guide on how to set it up to record tv programs. I record using an S-Video Cable, and thus far the results have been quite excellent. The followed his directions on how to record mpeg2 DVD files with MMC. The only thing I did different was I record at 720x480, and for some reason it gives me good results.

Now a great program for editing these files is called Womble Mpeg Video Wizard. It offers precise editing, right down to frame level. It doesn't re-encode unless absolutely necessary. If you are not changing the specs of your files, it doesn't re-encode at all. I think you realize that movies that have alot of action scenes, and are very well lit, will record bigger. Ones that are dark, and do not have all that much action record smaller.

I use Tmpgenc DVD Author 1.6 to design and author my DVD's. If you do end up with some that are slightly bigger than will fit of a blank DVD-R, you can authorize anyway and then use DVD Shrink to take this file and get it to the right size. I have heard that 90 percent or better is ideal, though I have heard some say that they can shrink down to 80 without that much of a loss in quality.

I suppose it wouldn't hurt to atleast try this method. I know that I am very happy with what my computer has been able to produce with this setup. Could you explain to me about how you are getting these files you are referring to as .m2v, I really don't understand how that is done, and why that is done? I hope I didn't break any rules, by asking that question.

[Tom Saurus]

BobaFett77: I didn't notice until I had already posted that you are living in Germany. I know different technology and broadcasting schemes are used in Europe. I kind of feel foolish now; but maybe some of what I wrote can be adapted to your situation. If you do a search for digitalfaq on the internet you should be able to find Lord Smurf's Website. He really has alot of valuable information there. I am very grateful for this resource he has provided on the Internet and I am also grateful for Videohelp.com.

[BobaFett77]

Generally, reencoding is better but slower that transcoding.
Among the transcoders I've tried (ReJig & DVDShrink) I've found no noticable difference in quality.

Is reencoding really better? I've read on some websites that transcode has better results, because it uses specific information from the already encoded MPEG stream and only changes some parts of it.

On this e.g. it says:

The goal of MPEG transcoding is to process one MPEG-compliant video stream into another MPEG-compliant video stream that has properties better suited for a particular application. Transcoding differs from the encoding and decoding processes in that both the input and output of the transcoder are MPEG video streams. A naive solution to the transcoding problem, shown in the top of Figure 1, involves the following steps: first, the MPEG-coded video stream is completely decompressed into its pixel-domain representation; this pixel-domain video is then processed with the appropriate operation; and finally the processed video is recompressed into a new MPEG video stream. Such solutions are computationally expensive and have large memory requirements. In addition, the quality of the coded video can deteriorate with each recoding cycle.

A more efficient transcoding solution, shown in the bottom of Figure 1, is to only partially decompress the video stream and perform the processing directly on the compressed-domain data. For some operations, further gains can be achieved by only processing the portions of the video stream that are affected by the transcoding operation. In this work, the affected portion of the video stream is partially decompressed into its motion vector (MV) and discrete cosine transform (DCT) representation. Computational requirements are reduced by eliminating the process of further decompressing the MV and DCT representation into the spatial domain, processing the spatial domain data, and recompressing the result into its new MV/DCT representation. For many compressed-domain transcoding operations, additional gains are achieved by exploiting the sparsity that is typical of quantized DCT coefficients.

From: http://www.hpl.hp.com/personal/Susie_Wee/PAPERS/hpidc97/hpidc97.html

Also take a look at this link: http://web.archive.org/web/20021223211323/http://www.bbc.co.uk/atlantic/transcoding.html

[BobaFett77]

Regarding my setup, as some people asked: I'm streaming TV programs with a satellite receiver with built in hard disk. It saves the program 1:1 as it is being broadcast by the TV stations as TS files (MPEG-2 transport streams). The resolution depends on the TV station, but most important stations use 720x576. I then transfer the TS files to the PC and demux them with ProjectX and receive separate video (m2v) and audio (mp2 or ac3) files. From those I author a DVD project then, which I burn on a DVD-R. If a recorded video file is too big for one DVD-R because of its length or bitrate I have to split it or shrink it though.

[Tom Saurus]

BobbaFett77: If possible can you download a free trial version of Mpeg Video Wizard and play around with it. It is capable of demuxing files. I wouldn't be surprised that it could handle these TS files that you mentioned.

I sometimes create 720x480 still pictures, in my photo editing program and do some pretty fancy lettering and pictures on the surface and then take this .bmp to Pinnacle Studio 9 ATI Edition, and use the make a title function. I select the picture as a background and don't bother with adding anything. I set a duration, then I select "make a movie" function and save the file as a 720x480 mpeg2 file. Since I haven't added any audio, it turns out as a m2v file.

Now when I am over in Mpeg Video Wizard, I find the folder that file is saved in, and select display all files, if I go with the default, display all supported files, it doesn't show that the mpv files are there. Well anyway I select the mpv file, and Mpeg Video Wizard accepts it. I put this on the time line and I add some audio to the audio time line and see that they match in length and then when I am exporting this Mpeg Video Wizard puts the files together and if I do it right, I end up with a DVD complient mpeg2 file. I only do this for making special titles for DVD's to introduce something really special.

I recently took a couple of still pictures from a video of a wedding my friend attended. I used Mpeg Video Wizard to capture that particular frame and I used my photo programs to edit the picture and placed it on a 720x480 blank picture I created. I actually did this with 2 pictures on 2 seperate 720x480 blank pictures. I know this sound awful convuluted. Well anyway I ran them through Pinacle, put the m2v files in Mpeg Video Wizard on the timeline. Then I put a transition effect which is another feature built into Mpeg Video Wizard between the two clips or two files I should say. I let Mpeg Video Wizard patch these two files together, with the transition, which was a "page turn" transition. Then I took the resulting file and started a new project. Then I went to the video of the wedding and I selected a small section where they were playing the wedding march. I found out exactly how long the duration of the video clip was, Mpeg Video Wizard is really good a providing this kind of information. I went to the clip of the video and I selected a portion with music that was this same length. I saved this clip and then I went to the demuxing tool and it extracted both an mpv and a mp2 file. I then created a new project and I took the mpv file and put it on the time line and then I took the mp2 file and put it on the sound line and the program patched them together into a DVD Compliant Mpeg Video file. It resulted in a very nice looking picture of some rich text, that announced the wedding, the wedding march playing in the background and then the page turn transition brought up the next picture which was a picture of the bride and groom and how "You are Cordially Invited To So and So's Wedding".

So in your case if 720x576 is the Pal DVD Spec's, I need to research that, since I am familiar with the NTSC Standard for DVD's; well anyway maybe Mpeg Video Wizard could do something with them. They have a nice tutorial in .pdf format that one can download over at Wombles Website and it has tons of information in it. Maybe there is something in that tutorial that would apply to TS Files. Actually in the Tools section where Mpeg Video Wizard is listed, if you scroll down to user comments, one person posted that Mpeg Video Wizard does a great job with PVR files. I wonder if perhaps, it might work for you. By downloaded a trial version of the program you have nothing to loose. I think you will find it to be so useful, that you will end up buying it. If you can atleast do some research on what Womble Mpeg Video Wizard is capable of, it might very well surprise you.

Sorry for going on and on. I hope perhaps you can find some helpful information in what I just posted.

[Tom Saurus]

BobaFett77: I got so caught up in recommending Mpeg Video Wizard, that I glossed over the fact that you already have something that is doing a good job at demuxing these files and that they are DVD compliant. I just got overexuberant, and I hope you can forgive me for that. But some of these shows must have commercials in them, or you need to trim them at the beginning or the end of the file. What do you use to do that? If you want something that will preserve as much of your file's original quality while editing, Mpeg Video Wizard is hands down the program for you. It is very quick as well. If nothing else atleast download the .pdf file at their website and it will describe the features that Womble Mpeg Video Wizard has, much more eloquently that I have. If you are just throwing stuff on DVD, commercials and all, you are wasting space on your media, even if you can fast forward through them at 20X speed. If you edit with some other program and it works well for you, stay with it ofcourse. Though sometime you might want to try Womble's Mpeg Video Wizard to see it compares well to what you have now. Don't worry I don't work for Womble, it is just that once I got the hang of that program it was just like going from a slow crawl to lighting fast when it came to editing out commercials. A matter fact when I bought Tmpeg DVD Author 1.6, I also bought their Mpeg Video Editor, and I never use it now; Womble just blew me away, once I understood how to work with it. Sorry for going on and on again.

[mats.hogberg]

I've read on some websites that transcode has better results

Well, that's wrong. Take a 1 GB mpg, transcode it to 500 MB and reencode it to 500 MB - The reencoded version will be better. The quote you provided doesn't say anything about the quality - just efficiency.

Such solutions are computationally expensive and have large memory requirements. In addition, the quality of the coded video can deteriorate with each recoding cycle.

True, mpeg reencoding (even at the same bitrate as the source) will detoriate the quality, but one reencode versus one transcode from the same source mpg, to the same dest size, transcoding will detoriate more than the reencoding does. I guess it's in the fiber of the universe - you can't make it both faster and better.

/Mats

[BobaFett77]

Well, at least in my case the video that was reencoded with TMPGEnc looked worse. Maybe I used the wrong settings, but comparing still frames from a close-up the reencoded version showed a clear loss of image details, while the same frame looked exactly the same as the original in the transcoded version. That was quite a surprise actually.

What program do you use for reencoding BTW?

[mats.hogberg]

I mostly use QuEnc, even if encoder shootouts use to find Canopus Procoder the overall best (if such a judgement can be objectively made). TMPGEnc use to be in the middle of the crowd.
But to be honest, 99.9% of the time, I'm quite satisfied with the result from transcoding.

/Mats