DivX:-; vs. SVCD

[Guest]

I find the recent two articles on DivX in comparison to MPEG2 relatively misleading. The statement that a 9-to-1 compression is even remotely likely is about as convincing that VHS copy of a DVD renders comparable results. I have spent considerable time both in the MPEG2 (SVCD) and DivX arena and found that for either to achieve high-quality results a datarate of about 2000mbits/sec is required. While DivX in this case does not require a reduction in resolution one should not overlook the fact that an excellently compressed 480x480 SVCD image beats a 720x480 image that exhibits macroblock noise, even at marginal levels. Considering the fact that DivX allows for cropping of the movie area the results are a tad disappointing. One should also note that DivX does not allow for VBR encoding and requires considerable computing time even for VKI (insertion at keyframes at scene changes through automatic detection). The DivX 3.2 codec simply doesn't cut it here and a two pass process is required, very time-consuming when using bicubic resizing.

The bottom line that I have found is that there are excellent commercial and free tools (TMPGEnc and Cinemacraft encoder) that allow for several pass VBR processing that in many cases let the final SVCD style image excel. The lack of standalone player support (DivX is a hack and will never gain commercial support) and the fact that it requires data CDs vs. SVCD CDs with reduced ECC, allowing more data to fit on a single CD, finally give SVCD (MPEG2) the advantage in the race for movie encoding.

Sorry, both articles in this area by Andreas Voelkel are a bit overenthusiastic for the MPEG4 format. The articles, in their comparison of quality also lack the insight that movie quality should not solely be judged upon a frame's appearance when rendered as a still but the whole impression of playback is to be considered. Many other areas were also insufficiently or plain incorrectly covered. Andreas could have done a better job but then again, this is a wide playing field and to come up with an interesting and concise article must have been hard.

Let's see what MPEG7 has to offer when it is discovered by the ripping community. Maybe then a real alternative to MPEG2 will emerge.

 

[lakedude]

I think you are being a little hard on Frank and his articles. DivX is certainly not perfect but it works and it is free. I use DivX on a 12' 400MHz laptop and on a 32' TV hooked up to a 1.3GHz desktop. The Laptop is too small to notice any minor microblock errors and most folks think it has a DVD player built in even though it doesn't. The 32' TV does show more flaws but it is hooked up to a much faster computer so the image quality is better hence you still do not see any microblock. The one really noticeable flaw is the reduction of color in large low contrast areas like sky, smoke or shadows. Color banding is noticable when there are large patches of colors without much contrast. Cartoons like South Park will show this flaw. If your puter is fast enough, microblock is not an issue. If your computer is slow SVCD may be a much better alternative but DivX is working good enough for me.

Also the fact that DivX is a hack is a plus not a minus for certain anti establishment types of folks.

Your post has sparked my intrest and I shall check out the SVCD stuff that is avalilable for free cause commercial stuff is not a fair contest.

<P ID="edit"><FONT SIZE=-1><EM>Edited by lakedude on 03/28/01 09:28 PM.</EM></FONT></P>

 

[Guest]

hi Revelator!

at first a small correction: the guys name is Frank not Andreas. but that's not the important thing.

you're are right that his articles about mpeg4 and mpeg2 are misleading and quite often simply wrong. most people who are a bit into the mpeg4/mpeg2 matter don't read his articles really. there've been some postings about this already on the german THG community board. unfortunatly, the unaware reader, new to the topic, is misguided by articles as stupid as these. that's a shame, I agree with you.
who says that a two hour movie looks the same either played from a DVD or from a CD (700MB coded to mpeg4) is blind or just lying.
I've done quite some mpeg4 coding from my DVD's now and I can tell the difference. of course, there are these people who now say I'm not doing it right. let them say.
I do like mpeg4. it does a good job for me and I'm quite happy with the quality/size ratio. I can happily live with some visible macroblocks in large low-contrast areas. this are most times not the areas where the action is and so your eyes won't rest on them anyway.
other thing for your info. mpeg2 also has macroblocks. and when you watch a DVD on a TV (or large monitor) you can see some macroblock noise. but it doesn't really matter since it's in low-contrast areas that are not the center of attention anyway.

respect!

 

[Guest]

Well,

what I tried to install was a drop of sanity here. MPEG4 is an excellent format and in its possibilities superior to MPEG2. However, the exploitation of this format is lagging many years behind MPEG2, a format that has matured, has been accepted by the community and offers even on the freeware level the arguably second best encoder in the world (TMPGEnc). What the current issue is that during the encoding process you can throw extra CPU cycles at MPEG2 and get bang for the buck. At some point MPEG4 simply maxxes out. There is no multipass VBR encoding (manipulating the bitrate allocation, not VKI) and that represents a big drawback.

In any case, both formats continue to deserve attention, just don't close the door on MPEG2 yet ...

/rev

 

[lakedude]

First let me conceed that you likely know much more about this topic than I. What are these errors you are talking about? Tom's page was the original source for my getting into MPEG4 coding and of course I have had to go elsewhere for more detailed info. Nothing I read on Tom's seems wrong just maybe a little optomistic. For the record you can definately tell a quality drop when going to MPEG4. The original Tom's artical shows the quality drop with a couple of images from the Matrix. It goes on to say that you hardly notice the difference which is only semi-true. As a still frame it is easy to see what I would call "JPEG" compression artifacts. If you are trying to see flaws you will. But as was already pointed out your eye would ordinarly follow the action not a large low contrast area. Well I'm off research SVCD so I don't gotta ask any real stupid questions. Thanks for bringing it up.

 

[Guest]

Just a note to say that DivX does support variable bitrate and it's called Smart Bitrate Control or SBC, and when used can create scenes in single DVD rips without macro blocks any bigger than a couple of pixels which you can only see if you like watching your movies with your head a foot from the screen.
But I back you up on the whole SVCD format thing, it's a brilliant standard which produces quite awesome results when done correctly

 

[Guest]

Well,

SBC and NanDub are a job IMHO. I personally do not hold a degree in rocket science to operate this program nor do I intend to diddle around with ominous parameters until hopefully someday I can create something useful.

SBC really only does two things and it does it in a very awkward way: I makes semi-intelligent decisions about whether to pick hihg- or lo-motion DivX for certain scenes and chooses places where to (according to its judgment, which again is dependent upon 101 parameters that you yourself have to try to figure out) set keyframes. While it is a two pass process it offers none of the advantages of a two-pass VBR MPEG process, i.e. floating adjustment of the bitrate.

Trust me, I am not a total idiot but I am also not interested in having to learn what gear ratios my car uses or exactly how its ignition control works in order to drive around the block. But that is a pretty fitting analogy in respect to use SBC. I prefer to turn the key and push the gas-pedal and I expect the thing to go.

SBC has years to go before it will turn into anything useful.

 

[Guest]

SBC actually does waaaay more than what you said. It goes far beyond the scope of low motion and high motion codecs and selects the compression routine for each frame just as in two pass VBR for MPEG1/2 in TMPEGEnc. It doesn't just choose the codec to use, in fact it uses the same codec for the entire film, instead it manipulates it at low level to allow far more control over the codec than just choosing high or low. Also recent updates to SBC has made the decision for the compression level of the frame much more advanced than in pervious versions.

NanDub is also pretty simple to use, all you do is load a profile for 1 or 2 CD rip and press shift F8

 

[Guest]

I don't think it NanDub is quite that simple to use. There is a forum dedicated to choosing profiles alone at Doom9, plus another one for SBC alone.

I wish it was that easy. Reading the NanDub guide at http/www.doom9.org/sbc_main.htm makes my head spin. While you are correct that the recent versions make it slightly easier (a few parameters became obsolete) it is still a substantial trial-and-error process.

If Doom's version is not the latest, let me know and I will gladly be educated.

As for Lo-Mo and Fast-Mo: You are incorrect. Nandub's 'Motion Detection' actually lets you set the thresholds as to when to use Lo-Mo and when Hi-mo.

As for complexity: There are no less than 37 different input settings for Nandub. I wouldn't call this 'Select 1 CD setting and hit F8'.

On top a setting of 'Anti-[-peep-]' (sic) somehow does not give me the peace of mind that the authors have reached the age of adult maturity <P ID="edit"><FONT SIZE=-1><EM>Edited by TheRevelator on 04/29/01 04:20 PM.</EM></FONT></P>

 

[Guest]

The SBC forums are for perfectionists, if you just want to get a decent rip all you need is a profile, a bitrate calculated so the rip will fit on 1 or 2 CD's and the go button, there is of course the audio muxing but if you can't handle that then you have no business talking about encoding in the first place