resolution loss in transcoding or rendering diagram needed

[Dwight Flynn]

This may sound like a silly statement but when it comes to how much resolution (on average) is lost when one transcodes or renders from one format to another I am usually lost. It would be helpful to have some sort of chart that allowed you to get these kinds of averages in a cross referenced matrix format; especially for the more popular formats and codecs (ie. avi, mpeg, mpeg4, divx, ts, m2t, mov, dv, etc.) I am putting this out there because as a tool for informed decisions, I think it would be invaluable, and I currently do not have the technical information to prepare it myself. Also, if it does exist in some accessible format, please share?

Thanks

[Glenn Chan]

Computer formats don't really lose resolution when transcoding, as resolution is related to the number of pixels and they can stay the same.

Generally the "generation loss" from compression/transcoding is measured by comparing the output vs the input on a pixel by pixel level. There's different ways to weigh the average difference. Also, your results will depend HIGHLY on the input (i.e. some codecs do a good job with cartoon animation, others with computer-generated stuff, while others do a good job with high motion material).

2- What are you trying to do?

[Dwight Flynn]

Glenn, given that there seem to be some categories that are more common than others (cartoons, computer generated, fast motion, etc) my basic question is still in play (using generation loss instead of resolution). I just want to know where I can find this information in readable format. I suggested a graph because it seems to lend itself to the format, but I am not wedded to the idea. If you or others know of a chart, series or charts, or a book with this information please let me know. BTW, I am trying to learn this stuff because I have the JVC hdv cam and access to other kinds (ie. dvx100, gl2, etc) and I find myself doing a lot of transcoding lately. I generally shoot docs and some drama shorts, so it is not fast paced shooting or animation (though that might change). I am curious to know how much image quality I am losing as I go from one to two to three generation in the different formats.

[Rob Lohman]

I don't see why anyone would actually find this helpful, besides
that, I don't think it is even possible to make such a chart. Here's
why:

1. it totally depends on the codec you are using (mpeg1, mpeg2 and mpeg4 are notorious for all the different implementations)

2. it can totally depend on the program that is feeding the encoder/codec

3. it can also depend on the version of the codec/encoder you are using

See how quickly things grow beyond your capacity to research
and make available in a correct and easy to read manner?

Just to follow your format list:

1. avi => is not an encoder/codec. It is a CONTAINER format. DV is a codec (for example) that you can use with AVI
2. mpeg => which version? I assume you mean MPEG1 (multiple encoders again)?
3. mpeg4 => that is the same as divx or xvid for example
4. divx => see point 3
5. ts => is MPEG2 (usually, doesn't have to be though!)
6. m2t => is MPEG2 (multiple encoders again)
7. mov => same as AVI, is a CONTAINER format. DV or Sorenson are codecs for this container format (for example)

It would be far better to test codec quality versus size and take
the market penetration into account.

However, you can already seriously cut down on the choices you
have depending on what you want to do:

1. output to (S)VHS => use DV and then your camera or a DV->analog converter

2. output to DVD => use MPEG2

3. output to the web => use MPEG1, WMV (windows media) or AVI/MOV with either mpeg4/divx/xvid or sorenson (in the case of quicktime)

So with option 1 you don't need to do anything. DV AVI is the
same thing as DV MOV (QuickTime) for example.

With option 2 you are either satisfied with your current MPEG2
encoder (make sure you LEARN what the various encoding options
do for best results!) or you are not and need to start looking at
a better encoder.

The last option is the hardest, but basically you only have 4
choices in my opinion (if you want lower resolution and thus
smaller filesizes):

1. mpeg1 => pretty large still and not looking to great, not used that much these days

2. WMV => windows media, it is also the codec

3. mpeg4/xvid/divx => in one form or another available for AVI or MOV (QuickTime). All basically do the same, but might differ a bit in quality. Some testing should easily identify which works for you

4. Sorenson => only works with QuickTime, is available in various versions and also in a pro version with more control

That's basically it. The only question you need to ask yourself is:

" is this looking good enough (or can it be done better) "

Basically. Ofcourse you want to make sure others can play your
file, and it isn't too large in filesize etc. But you get the idea.

[Dwight Flynn]

Thanks, Glenn/Rob. The problem I was having is that there are many of these formats, and I have tried many of them, but only a few are of any use for content delivery (ie. mpeg2, mpeg4, wmv. etc.) Even so, as you correctly pointed out, there are different variations of each, and I wanted to have some sort of quick reference for the few occasions that I have any sort of a choice of delivery formats. Your analysis was helpful, and I am going to do some more research.

[Michael Sinclair]

I have been experimenting with rendering my home movies to DVD. My captured DV looks bad when converted to MPEG2. It is annoying to see motion scenes because the fuzziness appears. I wish I knew that best codec for rendering to DVD from DV. Any ideas?

[Rob Lohman]

1. Canopus ProCoder
2. TMPGEnc / CCE (Cinema Craft Encoder)
3. Mainconcept (ie, Vegas & Premiere)

The big question is what are you using now Michael? DVD encoding
(as I've said often) is an artform as well and takes time to master,
understand what all the various options do. It is not a simple push
the button process as you've found out!

[Michael Sinclair]

I appreciate the advice. What I realized is that I should take a clip with motion that I know was part of a bigger DV transfer to my PC and just keep experimanting with that small clip. That will save space and time. My wife thought that I was being like the typical guy who takes the TV remote and goes crazy trying to balance the red, greens and blues. I showed her the DV film from our Canon Optura 300 directly hooked up to the TV and I then showed her the Mpeg2 conversion. She finally saw what I was talking about. The fuzziness that occurs. I also used S-video out from my PC video card and played captured DV to my TV. It looked the same as the Direct camera to TV clarity. The MPEG2 conversion clips played the same way when connecting the PC to the TV through S-video out on my Radeon 9800 XT. It looked fuzzy at times-- especially during motion. I wish that someone with more experience would list the programs and steps they use on a PC to get the best encoding to MPEG2. I know that it takes practice, but I would love to hear from PC users what their basic programs and approaches are to MPEG2 rendering from a DV format. I have been using Vegas by the way. I also took your advice and got the free TMP encoder. It pretty much looked the same as a Vegas render. I have Adobe Premiere Pro as well but I never used it yet. It looks complicated and I will buy some training DVD's for it.

[Rob Lohman]

Okay, then we get into settings (if you want get a demo of
Procoder Express as well from www.canopus.com).

What settings / encoding method did you use inside Vegas and
TMPGEnc? Also check the following threads where I posted a
heap of links to all sort of guides on the various TMPGEnc programs:

http://www.dvinfo.net/conf/showthrea...threadid=35635
http://www.dvinfo.net/conf/showthrea...threadid=35781

The main things to consider when doing an encode is:

1. your field order if you are doing an interlaced encode

2. the encoding method: constant bitrate (CBR) or variable bitrate (VBR, usually better if at least 2 passes, some people encode even at 9 passes for example)

3. the bitrate used: for constant bitrate this is one number which should be at least 5000 (5 mbit) and can go as high as 9800 or 9.8 mbps. VBR usually has three numbers, with minimum, average and maximum numbers. I worked with 0 - 7000 - 9000 in the past

[Dennis Vogel]

This article http://mediasoftware.sonypictures.co...m=&EntryPoint= discusses the parameters you can set with the MainCOncept MPEG-2 encoder used in Vegas.

Good luck.

Dennis

[Rhett Allen]

You absolutely have the ability to make this yourself. Just take a sample of your raw DV. Now compress it using all the various formats you have at your disposal and compare them.
There actually are a few sites with this sort of comparison on them, unfortunately I can't find one of them and the other one is offline (I do have it archived though). They are more comparing different video formats and one of them compares DV codecs (like Canopus, QuickTime, DVSoft, MS DV and so on). If you're worried about going to DVD then you need to do a bunch of testing to find what the best mpeg-2 encoder you can get is.
If you want to really be sick, dump a copy to VHS! I haven't watched VHS in so long that I recently made a dump on regular VHS (I usually use S-VHS for personal tape dumps) and was shocked at how horrible it looked next to the DVD of the same video. YIKES!

[Dwight Flynn]

Thanks guys, I'm going to slooowly build the information database I need through trial and error (with the help of a few handy web sites)