Sure way to compare picture quality...visual or bitrate? - AVS Forum | Home Theater Discussions And Reviews
Forum Jump: 
Closed Thread
 
Thread Tools
post #1 of 42 Old 01-09-2004, 06:13 PM - Thread Starter
Senior Member
 
pixel8's Avatar
 
Join Date: Jan 2003
Location: the Matrix
Posts: 321
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
I've read reviews so far and realized most were made out of personal visuals. Some reviews were in contrast and some were consistent.

However, wouldn't bitrates be the best way to compare quality for maybe same titles but different edition or different region? Isn't it consistent enough or is it not the deciding factor due to some other reason?

I've used Bitrate Viewer software to measure quite a few and it seems quite consistent. The ones that I visually liked were at an average of 8 and above eg. Jurassic Park superbit. LOTR-FOTR (EE) and TT (EE)was however below that and I kinda agree visually.

What do you think?

Clatto! Verata! Nnyaarrghahum(coughcough)!!
pixel8 is offline  
Sponsored Links
Advertisement
 
post #2 of 42 Old 01-09-2004, 06:15 PM
AVS Forum Special Member
 
FredProgGH's Avatar
 
Join Date: Mar 2002
Location: Chattanooga TN
Posts: 7,604
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 10
This is gonna be fun. Let's talk about DTS and partial birth abortion while we're at it :D :D

Don't taze me, bro!!
FredProgGH is offline  
post #3 of 42 Old 01-09-2004, 06:33 PM - Thread Starter
Senior Member
 
pixel8's Avatar
 
Join Date: Jan 2003
Location: the Matrix
Posts: 321
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
hehehe....huh? Not sure the relation between DTS and abortion but.....

I think throughout this forum, there have been many threads about THE reference picture quality in DVDs and of course many have replied, exhaustively. But, isn't there a true measurement for pq?

I'm not even going to dwell into sound quality 'cos I dun know any software which can at the moment.

Clatto! Verata! Nnyaarrghahum(coughcough)!!
pixel8 is offline  
 
post #4 of 42 Old 01-09-2004, 06:54 PM
AVS Forum Special Member
 
Kram Sacul's Avatar
 
Join Date: Sep 2003
Location: Los Angeles, CA
Posts: 5,558
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 10
There are plenty of spectacular dvds with an average bitrate below 6mb/sec. It's all about the telecine transfer quality and compression techniques.
Kram Sacul is offline  
post #5 of 42 Old 01-09-2004, 06:56 PM
AVS Forum Special Member
 
FredProgGH's Avatar
 
Join Date: Mar 2002
Location: Chattanooga TN
Posts: 7,604
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 10
Quote:
Originally posted by pixel8
hehehe....huh? Not sure the relation between DTS and abortion but.....

True.. abortion would probably be discussed more calmly around here :D

Don't taze me, bro!!
FredProgGH is offline  
post #6 of 42 Old 01-09-2004, 07:07 PM - Thread Starter
Senior Member
 
pixel8's Avatar
 
Join Date: Jan 2003
Location: the Matrix
Posts: 321
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Good point, Kram. If the source is bad, no matter how good the bitrate, pq will still be bad.

But how's the relation between compression technique with pq? Is it ....

bad technique -> good bitrate but bad pq?

Clatto! Verata! Nnyaarrghahum(coughcough)!!
pixel8 is offline  
post #7 of 42 Old 01-09-2004, 07:30 PM
Advanced Member
 
Cliff Olson's Avatar
 
Join Date: Dec 2003
Location: Petaluma, CA
Posts: 608
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 4 Post(s)
Liked: 10
Yeah, IMO, bitrate is almost meaningless. You can transfer crap at 10mbps, and it'll still look like crap. It's all in the transfer.
Cliff Olson is offline  
post #8 of 42 Old 01-09-2004, 07:38 PM
AVS Forum Special Member
 
FredProgGH's Avatar
 
Join Date: Mar 2002
Location: Chattanooga TN
Posts: 7,604
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 10
Well, here's the thing at least the way I understand it...
The bit-rate needs to be high enough on any given title to allow the encoding not to have visible artifacts. For a dark, simple scene this number can be theoretically low. For bright busy scenes it will probably be high. The bitrate doesn't mean much unless you know how difficult the scene in question was to compress. But, all things being equal if you keep the bitrate high you're less likely to run into trouble. The kicker is that most movies have scenes of varying complexity to decode, and you want to leave more bandwidth for the hard scenes and not waste it on scenes that don't require it. So, that's why the number fluctuates constantly throughout a DVD. It's dumb to waste a full 10Mb bitrate on the movie's titles when you might be hogging bandwidth that could have been used for a complex scene. And some movies seem to look great without requiring a large constant bitrate. Finding Nemo is one of these, no matter what anybody says. :D

Don't taze me, bro!!
FredProgGH is offline  
post #9 of 42 Old 01-09-2004, 07:47 PM - Thread Starter
Senior Member
 
pixel8's Avatar
 
Join Date: Jan 2003
Location: the Matrix
Posts: 321
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
For new movies, the transfers should be pristine though. So bitrates may not be applicable only on older movies.

Theoretically, measuring bitrates for new movies should be consistent, unless of course someone silly decides to downgrade or tarnish the source, right?

Clatto! Verata! Nnyaarrghahum(coughcough)!!
pixel8 is offline  
post #10 of 42 Old 01-09-2004, 07:59 PM
AVS Forum Club Gold
 
Art Sonneborn's Avatar
 
Join Date: Jun 2000
Location: Battle Creek,MI USA
Posts: 23,175
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 238 Post(s)
Liked: 295
I've seen this discussed about a year ago. I think the consensus then and what made sense is a bit starved transfer will not be a good as one given the minimum bit rate to not make that the limiting factor but a high bit rate is only one factor and surely doesn't ensure a reference PQ.

I think there are several anal retentive members who would have a pretty good consensus on what is and what constitutes a reference quality picture.

Art

My HT


iRule rules my theater
" Complaining makes you the victim.Change the situation, leave the situation or accept the situation, all else is madness"
Art Sonneborn is offline  
post #11 of 42 Old 01-09-2004, 08:02 PM - Thread Starter
Senior Member
 
pixel8's Avatar
 
Join Date: Jan 2003
Location: the Matrix
Posts: 321
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
FredProgGH, I agree with the ups and downs of bitrate throughout each movie. But if referring to average, some sort of standard can be revealed.

OK, let me try to put it in perspective .....

If bitrates is meant to be high on complex scenes, then movies like LOTR or Star Wars should register at almost 9 or 10 average bitrate. But if the average shows less, then that must mean something... maybe disc space limitation due to special features, soundtracks, etc. I'm assuming superbits and regular releases are from the same sources but the difference is more disc space to utilize for superbits. Hence, more bitrate for superbit.?

Furthermore, if comparing the same movie between regions, different bitrates may mean better pq (if same source) or subjective pq (if different source).

Clatto! Verata! Nnyaarrghahum(coughcough)!!
pixel8 is offline  
post #12 of 42 Old 01-09-2004, 08:22 PM
AVS Forum Special Member
 
FredProgGH's Avatar
 
Join Date: Mar 2002
Location: Chattanooga TN
Posts: 7,604
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 10
If the bitrate is low, it does mean something, but what? Does it mean they would have liked to have had a higher rate, but couldn't because the bit-budget was tapped out, or a higher rate wasn't necessary so they were able to have more content??

It's all an art- telecine, compression, authoring. If the results look good, I don't sweat it. Of course, I do agree that a you can manage to keep the bitrate high, it means you are potentially allowing a higher quality picture. What Art said is basically true I think.

Don't taze me, bro!!
FredProgGH is offline  
post #13 of 42 Old 01-09-2004, 08:22 PM - Thread Starter
Senior Member
 
pixel8's Avatar
 
Join Date: Jan 2003
Location: the Matrix
Posts: 321
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Few more cents ....

Especially these days, we have multiple regions/releases/editions of the same movie eg. Alien series, Terminator series, LOTR series, etc. How can one determine for sure the better pq?

Looking at the 1st page of this forum, there's bound to be a thread asking for reference dvd/frame/demo in terms of pq and sq. Each response would be different from each member. So, how does one mean by 'reference'?

Clatto! Verata! Nnyaarrghahum(coughcough)!!
pixel8 is offline  
post #14 of 42 Old 01-09-2004, 08:23 PM
AVS Forum Special Member
 
FredProgGH's Avatar
 
Join Date: Mar 2002
Location: Chattanooga TN
Posts: 7,604
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 10
FWIW, many of the Superbit titles clearly demonstrate the effect increasing the bitrate can have on a given transfer. And a couple of them demonstrate the lack thereof...

Don't taze me, bro!!
FredProgGH is offline  
post #15 of 42 Old 01-09-2004, 08:33 PM
AVS Forum Special Member
 
FredProgGH's Avatar
 
Join Date: Mar 2002
Location: Chattanooga TN
Posts: 7,604
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 10
I remember having a "What do you mean by reference" thread a while back...
I define it as "having the least deviation from the source material." To a large extent you can have an objective analysis of the type that Bjorn Roy excels at, and beyond that it becomes a subjective "what I like" kind of thing.

Don't taze me, bro!!
FredProgGH is offline  
post #16 of 42 Old 01-09-2004, 09:16 PM - Thread Starter
Senior Member
 
pixel8's Avatar
 
Join Date: Jan 2003
Location: the Matrix
Posts: 321
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Ok, conclusion is .... bitrate measurement is accurate if same source. But for different new movies (different source), bitrates can be used as a benchmark.

FredProgGH, you mentioned Finding Nemo having low bitrate but good pq .... do you know the average bitrate?

Clatto! Verata! Nnyaarrghahum(coughcough)!!
pixel8 is offline  
post #17 of 42 Old 01-09-2004, 09:19 PM
AVS Forum Special Member
 
FredProgGH's Avatar
 
Join Date: Mar 2002
Location: Chattanooga TN
Posts: 7,604
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 10
I haven't checked it myself but I think people have reported it averaging in the 3.5 to 4 range. I think they got away with it because (as mentioned ad nauseum in another thread) Nemo has an intentional soft, diffused look in most of its scenes and maybe wasn't as hard to compress as some other Pixar titles. The results are still great!

Don't taze me, bro!!
FredProgGH is offline  
post #18 of 42 Old 01-09-2004, 09:28 PM
AVS Forum Addicted Member
 
Dean Roddey's Avatar
 
Join Date: Aug 1999
Posts: 19,846
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 388 Post(s)
Liked: 293
All animation is easy to compress. One big eater of bits is the need to represent the complex textures that make up most natural physical objects. These introduce lots of high frequency detail that requires more data to store, even if a lot of it is lost due to the DCT type compression used by MPEG2. Animation usually doesn't have this problem, because there are large (relatively speaking) swaths of single colors, or gradual gradients, that compress very well.

Dean Roddey
Chairman/CTO, Charmed Quark Systems, Ltd

www.charmedquark.com

 

Dean Roddey is offline  
post #19 of 42 Old 01-09-2004, 09:57 PM
AVS Forum Special Member
 
FredProgGH's Avatar
 
Join Date: Mar 2002
Location: Chattanooga TN
Posts: 7,604
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 10
Quote:
Originally posted by Dean Roddey
All animation is easy to compress. One big eater of bits is the need to represent the complex textures that make up most natural physical objects. These introduce lots of high frequency detail that requires more data to store, even if a lot of it is lost due to the DCT type compression used by MPEG2. Animation usually doesn't have this problem, because there are large (relatively speaking) swaths of single colors, or gradual gradients, that compress very well.
Generally speaking, but I think less so with 3D computer animation vs. traditional 2D because computer generated animation can allow for much more complex shading, texture maps, lighting effects and so on. I mean, think of Final Fantasy, on one hand. Most Pixar stuff probably resides squarely in the middle, but I think even though FN has a lot of detail its diffused nature does place it closer to simple 2D in terms of its compression needs. Probably waaaaay more than Monsters, Inc. with all that fur to deal with!!

Don't taze me, bro!!
FredProgGH is offline  
post #20 of 42 Old 01-10-2004, 05:54 PM
AVS Forum Addicted Member
 
Josh Z's Avatar
 
Join Date: Jul 2002
Location: Planet Boston, source of the spice, Melange.
Posts: 22,543
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 2137 Post(s)
Liked: 1434
The bit-rate is just a number. The number itself is meaningless. What's important is the quality of the compression work. This is something that cannot be mathematically measured. Analysis of the results is entirely subjective.

Josh Z
Writer/Editor, High-Def Digest (Blog updated daily!)
Curator, Laserdisc Forever

My opinions are my own, and do not necessarily reflect those of my employers.

Josh Z is online now  
post #21 of 42 Old 01-10-2004, 07:00 PM
AVS Forum Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,386
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 107 Post(s)
Liked: 56
Quote:
Originally posted by Josh Z
What's important is the quality of the compression work. This is something that cannot be mathematically measured. Analysis of the results is entirely subjective.
I'm afraid that's incorrect. Compression quality can be measured quite easily.
All you have to do is compare the original pixel values to the reconstructed
pixel values. This metric is called PSNR (Peak Signal to Noise Ratio). See:

http://www.vsofts.com/codec/codec_psnr.html

Although PSNR is not the perfect metric of compression quality (because it
tries to assign one number to an entire image), it's what everybody uses to
tune their encoder algorithms.

Ron

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is online now  
post #22 of 42 Old 01-10-2004, 10:32 PM
AVS Forum Club Gold
 
Art Sonneborn's Avatar
 
Join Date: Jun 2000
Location: Battle Creek,MI USA
Posts: 23,175
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 238 Post(s)
Liked: 295
Quote:
Originally posted by pixel8
Few more cents ....

Especially these days, we have multiple regions/releases/editions of the same movie eg. Alien series, Terminator series, LOTR series, etc. How can one determine for sure the better pq?

Looking at the 1st page of this forum, there's bound to be a thread asking for reference dvd/frame/demo in terms of pq and sq. Each response would be different from each member. So, how does one mean by 'reference'?
It's like describing exactly what constitutes beauty, tough to do but few disagree when they see it. I would say in my collection only about one in ten are reference maybe two in ten are close. I'll bet most of the videophiles would agree on most of my choices.

Art

My HT


iRule rules my theater
" Complaining makes you the victim.Change the situation, leave the situation or accept the situation, all else is madness"
Art Sonneborn is offline  
post #23 of 42 Old 01-10-2004, 10:53 PM
Member
 
gabe nell's Avatar
 
Join Date: May 2001
Location: Seattle, WA
Posts: 156
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Bit-rate certainly is not a good way to try to quantify image quality. The same encoder could encode for the same target bit rate and spend different amounts of effort doing the motion estimation, for instance, and come up with visibly different results.

dr1394 is correct that PSNR is the defacto objective measure in video compression.

PSNR is used for several reasons: it is mathematically tractable, it is reasonably correlated to subjective 'measurements', and there aren't any real alternatives. Improving objective quality measure is an open research topic.

Now, if the goal is to take a DVD and try to objectively measure it's quality, PSNR won't do you a lick of good (and I don't think any objective measure could), since (presumably) you don't have the uncompressed source. :)
gabe nell is offline  
post #24 of 42 Old 01-11-2004, 12:17 AM
AVS Forum Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,386
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 107 Post(s)
Liked: 56
Quote:
Now, if the goal is to take a DVD and try to objectively measure it's
quality, PSNR won't do you a lick of good (and I don't think any objective
measure could), since (presumably) you don't have the uncompressed source.
That right, you can't do PSNR measurements without the uncompressed source.
However, there is a metric that can be extracted from the bitstream, the
average quantization level (mquant) of the macroblocks of a frame.

Here's a graph of a sequence in "Finding Nemo" charting average bitrate and
and average mquant per frame. As you can see, the bitrate changes quite a
bit, but the average mquant stays fairly well contained. This is the goal
of VBR encoding, to adjust the bitrate to maintain constant quality.

Ron
LL

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is online now  
post #25 of 42 Old 01-11-2004, 12:19 AM
AVS Forum Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,386
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 107 Post(s)
Liked: 56
To correlate the previous graph, here's a frame from the beginning.

Ron
LL

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is online now  
post #26 of 42 Old 01-11-2004, 12:20 AM
AVS Forum Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,386
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 107 Post(s)
Liked: 56
And the end.

Ron
LL

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is online now  
post #27 of 42 Old 01-11-2004, 12:23 AM
AVS Forum Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,386
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 107 Post(s)
Liked: 56
Here's the frame from that sequence with the highest average mquant (at 5.4).

Ron
LL

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is online now  
post #28 of 42 Old 01-11-2004, 12:26 AM
AVS Forum Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,386
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 107 Post(s)
Liked: 56
Here's the mquant per macroblock of that frame. The heavy motion of the
wing and head is difficult to encode and get's more quantization. "Hotter"
colors are higher quantization levels, with the pink blocks at 10 to 11.

Ron
LL

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is online now  
post #29 of 42 Old 01-11-2004, 08:36 AM
kal
AVS Forum Special Member
 
kal's Avatar
 
Join Date: Jul 2000
Location: Ottawa, Ontario, Canada
Posts: 7,187
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 16 Post(s)
Liked: 66
Quote:
Originally posted by Josh Z
The bit-rate is just a number. The number itself is meaningless. What's important is the quality of the compression work. This is something that cannot be mathematically measured. Analysis of the results is entirely subjective.
Josh hit the nail on the head. To determine picture quality you really have to LOOK at the result with your EYES. There's no mathematical way to determine picture quality. Just like in the audio world, no matter how many facts and figures manufacturer's throw at us, the most important thing you can do to judge audio quality is to simply LISTEN. Everything else is meaningless.

Kal
kal is offline  
post #30 of 42 Old 01-11-2004, 09:36 PM
Member
 
gabe nell's Avatar
 
Join Date: May 2001
Location: Seattle, WA
Posts: 156
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Ron

Nice examples.

Bitrate and mquant in Nemo are less correlated than I would've guessed, according to your graph.

Now, what I don't understand ... usually mquant refers to the step size, so shouldn't higher mquant be a coarser quantization, using fewer bits?
gabe nell is offline  
Sponsored Links
Advertisement
 
Closed Thread Movies, Concerts, and Music Discussion

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off