Lucky Number Slevin and CCE OPV

[nashcity]

I just encoded Lucky Number Slevin (good flick btw) using CCE OPV and I found it interesting that I was able to achieve a Q-value of 1 on the main feature at about 70% compression. I use OPV quite a bit and this is quite rare. I checked the quality and its amazing, it looks exactly like the original, and I'm very picky. This allowed me to keep alot more on the disc than I was originally going to (deleted scenes AND making of). I remember Fight Club being a similar case, where I was able to keep a very low Q-value at about 60% compression.

Just thought I'd pass that along. Am I correct in assuming that a Q-value of 1 (or a very low # in general) should be top notch in terms of quality, no matter what the compression? My results tell me YES. Could there be a situation where a low Q-value wouldn't be ideal?

[jdobbs]

A Q Value of 1 would indicate the highest quality level achievable. A low Q is never a bad thing. Compression levels, on the other hand, tell you very little about quality. That's because the original compression level, source complexity, and bitrate also have to be considered. It's like saying "what's better 50% of a dollar or 99% of a dime".

[nashcity]

Thanks for the info. I personally think OPV is a great way to discover just that: what kind of source are you working with? I would sure like to know what I'm working with before making decisions on compression levels and which features to keep/strip. It may take a little more management (vs 1 click and forget it), but in the end its something that I am more than willing to do. After all, this is supposed to be fun isn't it :)

No matter what I decide, I usually run the OPV prepare phase just to see what I working with. I may very well decide to use multipass after that (of course running another prepare stage in doing so), but at least it gives me a better idea of what compression ranges I want to be in.

[windtrader]

It's like saying "what's better 50% of a dollar or 99% of a dime".Shoot. By my definition I'd take 50% of a buck everytime. :o

A Q under 12 or so is very good (no or very minimal visible degradation), Q 12-24 - good (very minimal to slight degradation), and 24+ (likely visible degradation).

That's because the original compression level, source complexity, and bitrate also have to be considered
Additionally, the quality of the master input to the primary DVD encoding process is a key factor as well. What goes in may be several generations removed from the original acetate. Also, equipment (AD, DD, encoding) and experience of encoder can make a huge difference on overall quality. I guess generally GIGO (garbage in, garage out). However, it is amazing what AVISYNTH filters can do to improve the look of some movies.

[jdobbs]

After all, this is supposed to be fun isn't it :)Absolutely. :cool:

[manono]

A Q under 12 or so is very good (no or very minimal visible degradation), Q 12-24 - good (very minimal to slight degradation), and 24+ (likely visible degradation).

Well, I don't buy that at all. High bitrate matrices can stand higher level Qs before artifacting than can lower bitrate matrices, so as an iron-clad rule it's flawed from the get-go. But let's assume you meant those values are OK for CCE's default Standard Matrix. Are you aware that in the VBR Bitrate Allocation Window, the Q-Scale turns from green to grey at Q 7.5, and from grey to red at 15? Long before Q 12 you'll get at least some artifacting in the form of a softened picture and increased mosquito noise, and way before 24 you'll get additional and serious artifacting, such as color smearing and macroblocking. There is no way in the world that an average Q of 20 can in any way be called "good". It may look OK on a 19" interlaced CRT TV set, but not on any decent display.

Nor do I subscribe to the theory that the lower the Q the better, in all circumstances. My opinion is that if you got a Q of 1 or close to it, you probably haven't done all you could to tweak the encode and haven't taken full advantage of MPEG-2's compression abilities. Tweak in the form of a much better quantisation matrix, or the Standard Matrix with AQ turned on. And this assumes you haven't softened the encode, and thus lowered the Q "artificially" by using AviSynth denoising/smoothing filters, or CCE's low-pass or vertical filters (the low-pass filter being on by default). It also assumes a top quality source, such as many of the better Hollywood film DVDs.

I realize this flies in the face of the prevailing opinion around here, but I'm putting it out just so you'll know there are other opinions on the subject.

[jdobbs]

Agreed. Q is like every other value. You have to consider it within a given, defined set. Of course if I used a filter that made the picture black -- even though I'd get a very low Q --that doesn't necessarily mean the loss of detail improved the picture. The same is true if I resized it to Half-D1... the Q would be lower, but the resolution was halved.

But... I do think it is safe to say that within a given set of parameters (matrix, resolution, source, etc.) lower Q always means better picture.

[manono]

Can't argue with that. :)

[windtrader]

Humm.. maybe you had an off day but I did not make any kind of claim that was iron-clad. My comment, taken in context to the thread posts, is intended as a rule-of-thumb of Q values and general indication of visible quality of an encode vis-a-vis the original DVD. The context is already established there are many variables that make predicting or indicating encoding quality very difficult.

the Q-Scale turns from green to grey at Q 7.5, and from grey to red at 15

But... I do think it is safe to say that within a given set of parameters (matrix, resolution, source, etc.) lower Q always means better picture.
Given my premise, you can not establish iron-clad rules on Q and quality, color indicators are moot as well. As presented here, this subject is full of exceptions but after re-encoding well in excess of 300-400 commercial DVD titles, I can say that plenty of the encodes with Q < 15 exhibit far fewer visible artifcats than the other batch. This is viewing on an 8" FP CRT thrown onto a 100" screen.

And lower Q usually indicates better quality. Take a randon sample of 20 movie titles larger than a single DVD capacity and of various lengths. Encode them to fit on a single DVD-R, then group them into a below 15 Q and above 15 Q pile. I'd bet the lower Q batch will have better visible qualtiy than those with the higher Q values.