DVD-RB PRO RC6, AQE 0.31b2 and CCE 2.67: Results, analyses and comparisons.

[danpos]

Presentation:

Source: "The Day After Tomorrow" (2004) - 720 x 480 / 16:9 / 124 minutes.

Tests: FULL BACKUP (in truth, I got rid of spanish audio and subtitle  ) using DVD-RB PRO RC6. In first test was used AQE 0.31 beta2 and in the second one was used CCE 2.67 / ECLCCE 1.81.

Purpose: Observe the AQE behaviour when is selected ONE-CLICK / OPV with Q-prediction. We don't must to expected good accuracy with this mode cos the Q-prediction routine was thought for CCE encoder, not for AQE one (note: AQE is used like a 'CCE version' - RB until reads out the version from AQE's header and reports it in log).

Results:

Log's files: RB's log - AQE 0.31 beta2 /// RB's log - CCE 2.67/ECLCCE 1.81

Encoding's time (from logs): AQE = 100 minutes /// CCE = 74 minutes. 

Sizes (converted DVDs): AQE = 3.17 GB /// CCE = 4.18  GB. 

Comments: As I already there was said, the Q-prediction routine in RB wasn't working for AQE. I'm almost sure that undersize caught up by AQE use is due Q-prediction one. I've based this hipothesys on my results available at following links:  link 1 - Full encoding with Q-prediction and link 2 - comparing AQE with CCE in using same source, ECL file and Q-value. Perhaps the reason be other than last one. I'll do a FULL encoding 'by hand' and I'll use my Q-prediction method and so I'll se what I get at end ASAP. By the way, I've burned a DVD-R with DVD-RB/AQE compilation and I'll watch it in my SAP soon in order to test it out on playback consistence. Soon I'll report the results.

Screenshots:

AQE

CCE

AQE

CCE

AQE

CCE

AQE

CCE

AQE

CCE

Comments

Cya!

[techreactor]

I can notice the following :

-AQE is making the picture darker
- Some details might get lost due to the added darkness
- Is Taking more time
- Lesser file size.

[SAPSTAR]

Hmmm I would say don't use the OPV mode of DVDRB.. The sizing is very bad with it....I know that jdobbs is still working on it. FishMan0919 had better results with DVDRB + AutoQ or another tool more specialized.
Anyway, the OPV is rather "new" as I'm not a fan of this method, as I always say : better slow and good than fast and bad !! And women wont say the contrary ! 

@all : I will start working harder on the OPV mode from now as you guys seem to use it, I owe you that.....

@techreactor : did you really find the picture darker ?! Could be the consequence of a bigger compression (3.17/4.18 ...big diff...)

[danpos]

@ALL

One brief addition: the correpondent frames on differents compilations don't correspond the same KIND OF FRAMES because each encoder has a self way to do the process. For instance, the two last pictures: AQE picture is a B-FRAME whereas CCE one is a I-FRAME. I chose these two pictures because they show the difference between each one clearly (of course the I-frame will keep more details than a B-frame).

@techreactor

THX for your feedback. Regards to encoder's speed, I will wait that CCE were "the winner" because it's the faster encoder at moment. I did see this point at another way: 26 min of diference to CCE time encoding is very good e shows that AQE is a fast encoder too.  Regards to darkness from AQE's pictures, I don't see this ...  Perhaps a I need glasses for it.   

@SAPSTAR

This was the first time that I used DVD-RB in OPV mode. I always use 2 pass VBR in my RB's backup. I just do it in order to test AQE out in several scenarios.  BTW I did like from time encoding and with the due tweaks I'll can to switch to this mode in the future. 

C u all soon.

[danpos]

@All

Another additional clarifying: if you took a look at logs, you could see:

AQE used Q=38 for VTS_01 whereas CCE used Q=12 for same VTS.

AQE used Q=161 for VTS_04 whereas CCE used Q=42 for same VTS.

These can to explain the big difference found. Again in this post I showed that the same movie (a piece of VOB) encoded with both AQE and CCE and using the same settings (same ECL file): both reach the same file size!! 

C u soon.

[SAPSTAR]

@All

Another additional clarifying: if you took a look at logs, you could see:

AQE used Q=38 for VTS_01 whereas CCE used Q=12 for same VTS.

AQE used Q=161 for VTS_04 whereas CCE used Q=42 for same VTS.

These can to explain the big difference found. Again in this post I showed that the same movie (a piece of VOB) encoded with both AQE and CCE and using the same settings (same ECL file): both reach the same file size!! 

C u soon.

Thank you for this clarification Danilo !!! It demonstrates that the Q selection mode from DVDRB is not that good...Sometimes it gets totally lost....
Anyway, I'm to work on the OPV mode, FYI I just added a new algo in the QMatOp part ...it's the EDS algo, it's normally used in radar imaging to normalize the matrices. I implemented it to normalize the final matrices.....Once I will have finished tetsing it, I will try to improve the OPV mode....

cya !!

[danpos]

Anyway, I'm to work on the OPV mode, FYI I just added a new algo in the QMatOp part ...it's the EDS algo, it's normally used in radar imaging to normalize the matrices. I implemented it to normalize the final matrices.....Once I will have finished tetsing it, I will try to improve the OPV mode....

cya !!

WOW, this is a gret news !! I forgot to comment that I did observe that QMatOp in the new versions are more faster than previous (keeping the good results of course). BTW, I did some 'googling' and I found this paper out about EDS algo. It seems interesting and I will read it out soon (just for cultural purpose ).

Kept up, man! 

[SAPSTAR]

...I forgot to comment that I did observe that QMatOp in the new versions are more faster than previous (keeping the good results of course). BTW, I did some 'googling' and I found this paper out about EDS algo. It seems interesting and I will read it out soon (just for cultural purpose ).

Kept up, man! 

You're absolutly right my friend ! I modified QMatOp to be faster with higher bitrates(>3750)kbps...the counter part is it is slower (but more precise than before) for lower bitrates (<2500kbps)

[danpos]

You're absolutly right my friend ! I modified QMatOp to be faster with higher bitrates(>3750)kbps...the counter part is it is slower (but more precise than before) for lower bitrates (<2500kbps)

It's good to know about that. I realize that this is the correct way becaus as you already there were said for higher bitrates a flat matrix close to MPEG Standard can to give good results, whereas for lower bitrates the matrix must to be optimized for compression with quality (I mean to keep how details as possible .

Good point.

Cya!

[SAPSTAR]

You're absolutly right my friend ! I modified QMatOp to be faster with higher bitrates(>3750)kbps...the counter part is it is slower (but more precise than before) for lower bitrates (<2500kbps)

It's good to know about that. I realize that this is the correct way becaus as you already there were said for higher bitrates a flat matrix close to MPEG Standard can to give good results, whereas for lower bitrates the matrix must to be optimized for compression with quality (I mean to keep how details as possible .

Good point.

Cya!

Exactly...BTW for your culture you can read : http://www.measurement.sk/2004/S3/Mastriani.pdf

I used this document to create my EDS proc for matrices....It's smoothing the matrices values thus it reduces the "too big" steps, increasing the visual quality...

[danpos]

@SAPSTAR

THX mate! I will read it out asap.

Cya!

[Zyphon]

@danpos & techreactor

Thank you guys for your input and testing that you have done with this encoder.

Danilo those screenshots looks really cool and show clearly the differences between the two encoders.

@SAPSTAR

Thanks for all your hard work in developing this excellent encoder and the improvements that you have made it is getting better all the time.

I wish I could offer to do some tests also but I recently lost my hard disk to a suspected virus and I lost all of my HDD parition details and files therein. I am a bit depressed at the moment and need to find a way to retrieve some data from the hard disk. 

Until I can sort out my hard disk probs I will be using  my laptop just to keep track of this very, very interesting thread.

Keep up the great work guys.

[Prodater64]

Until I can sort out my hard disk probs I will be using  my laptop just to keep track of this very, very interesting thread.

Keep up the great work guys.

OT
I asked you before when you posted that problem.
Did you to recover your HD a very short time ago?
If not search something about recover HD in kvcd.net something not related to video. 1 or 2 months ago.
END OT

[danpos]

@Zyphon

Thanks for yours kinda words. By the other side I'm sorry with the happened with your HDD as I know that you're a enthusiastic guy that have the open mind for news about encoders and related stuffs and also for test them out as well. I hope that you can to recover yours dates and enthusiasm.

Up man!

Cya!

[Zyphon]

@Prodater

Hi Luis, yes I recovered from that problem.  That problem was caused by Parition Magic messing up on a drive resize and it messed up my partition.

I recovered from that one, this one seems more serious, I had XP instald on 3 partitions for seperate things and I lost all but one of those partitions and the one remaining wont load, my partition table is messed up.

Hopefully I can use the same tool I used last time to try and recover data form it. >>This<< tool.

@danpos

Thanks for the kind words Danilo, I guess I will be out of video encoding for a couple of months until I can buy a new HDD.

Major appologies guys for this off-topic post, I just wanted to reply to Pro's and danpos's post.