Great encoding with DVDrb-Pro 1.02 & AutoQmatEnc

[Sunhillow]

I just had a quick look at SW Episode 3, encoded with DVDrebuilder pro 1.02 and AQE.
It is Film only, 2 angles of view, german & english audio, no additional filters. It is really looking great, very sharp and no blocks visible even on explosions or high motion scenes. Switching between angles works fine.
And the best is, that my DVD player (Pioneer DV-444) did not stutter a single time as it always happened with older versions of HC encoder. It seems to be very picky with GOP length and max. bitrate.

Thanks for these great tools, jdobbs and sapstar!
:thanks:

[masscamp24]

What Qmatrix settings in RB-PRO v1.02 did you use.

[rendez2k]

Just a quick question about AutoQMatEnc - I'm just giving the encoder its first test, and I've enabled its 'QMatOp technology' (it looks like the encodes going to take about 12 hours to do a 2-pass encode!) - but what exactly does it do? Does it analyze each scene then apply a different matrix as required? Does all that extra time make a big difference (hopefully I’ll be able to see this myself in about 6 hours time)?!

[Fishman0919]

Just a quick question about AutoQMatEnc - I'm just giving the encoder its first test, and I've enabled its 'QMatOp technology' (it looks like the encodes going to take about 12 hours to do a 2-pass encode!) - but what exactly does it do? Does it analyze each scene then apply a different matrix as required? Does all that extra time make a big difference (hopefully I’ll be able to see this myself in about 6 hours time)?!


I would set AQE to 5 or 6 passes, if AQE get the right size in 2 passes... it will skip the rest.

Does it analyze each scene then apply a different matrix as required?

Yes, it checks each segment it encodes in DVD-RB and applies the correct matrix for that part

[rendez2k]

I would set AQE to 5 or 6 passes, if AQE get the right size in 2 passes... it will skip the rest.

Arh, so even the number of passes is variable? Very interesting. By 'get the right size' I take it to mean the best quality AutoQmatEnc can do or does it mean closer to DVD-R size?
I notice it doesn't get much of a mention as some of the other encoders (or even have default support in DVD-RB). Has anybody done any quality tests?

[Fishman0919]

By 'get the right size' I take it to mean the best quality AutoQmatEnc can do or does it mean closer to DVD-R size?

Will, say the target size is 100,000k for a segment to be encoded and on the 2nd pass AQE hit 100,001k. It will skip the remaining passes because the target size is just about right. Passes in AQE are not for Quality but for sizing.

Quality can sometimes be on par with HC and CCE but it's still being test and SAPSTAR is always tweaking the rate control and motion Algo's

[danpos]

I notice it doesn't get much of a mention as some of the other encoders (or even have default support in DVD-RB). Has anybody done any quality tests?

An interesting thinking. AQE (AutoQMatEnc) is relatively 'new' encoder and not many people had interest to test it out (I know that here at D9 members - Carpo, Axlemar, Fishman0919 and some others - have test AQE out since early versions). I thinking that how more people experiment it out and so report yours good results, more people will have interest (and curiosity) to do its tests itself and this will increase as a exp law. In this way, it will certainly mentioned by more people and perhaps be integrated in third-party programs like DVD-RB and others (for now already is possible to use it on 'all-in-one' DIKO application).

Regard on your question about quality tests, you can see several test reports done by several beta-testers at AQE's official forum (http://www.vmesquita.com/forum/index.php?board=41.0). What I can say as a very active beta-tester is that the last release is very reliable for multipass encodings (several tests performed have showed that 1 VAF + 2 PASS VBR is so far enough). Some tests that I've ran show me that bitrate spikes were gone. Also, I've working on OPV mode a lot and having good results (with some others not so good). SAPSTAR is currently working on this mode in order to improve it and soon it will be more reliable.

AQE is on continuous development and this is very good thing for yours users.

Last but not least, do yours tests yourself and see what you get. ;)

Greetings,

[rendez2k]

I'll let my 2-pass test complete, then I'll have a look at the results. Just out of interest, what is the best bit of software to take screenshots of videos - ideally something you can input time values into so you always get the same point and something that takes images at the correct ratio!?

[Sunhillow]

What Qmatrix settings in RB-PRO v1.02 did you use.

I used CCEAQM=1 in the REBUILDER.INI, so AQE calculated the matrices by itself.
One disadvantage of this encoder is its slowness, but the quality was really good for a compression of ~63% IIRC.

[masscamp24]

I used CCEAQM=1 in the REBUILDER.INI, so AQE calculated the matrices by itself.
One disadvantage of this encoder is its slowness, but the quality was really good for a compression of ~63% IIRC.

Thanks. However can AQE work with the Qmatop off and produce the same result with it on as encoding with it on takes a very long time.

[raquete]

Yes, it checks each segment it encodes in DVD-RB and applies the correct matrix for that parti don't know dvd-rb and AutoQMatEnc but this is interesting.
to understand how it work please answer me
1 what matrices are used and how the correct matrix is selected?i mean, who or what define that correct matrix?
2 what parameter observable to apply that matrix?
3 have general consensus to apply that correct matrix choosed in the segment checked?

thank you.

[danpos]

i don't know dvd-rb and AutoQMatEnc but this is interesting.
to understand how it work please answer me
1 what matrices are used and how the correct matrix is selected?i mean, who or what define that correct matrix?
2 what parameter observable to apply that matrix?
3 have general consensus to apply that correct matrix choosed in the segment checked?

thank you.

1. The 'built-in' QMatOp algorithm present in AQE. It uses the DCTune (by NASA) for scanning several GOPs in order to find out the matrix which will represent the more adequate cost x benefits (compression x quality) for the desired target bitrate. More about QMatOp will can see at <here> (http://qmatop.9k.com/main.html).

2. This is controlled by DCTune (http://vision.arc.nasa.gov/dctune/) and more some others tweaks which SAPSTAR has done. Follow the link present for more informations about DCTune. ;)

3. The matrices generated by QMatOp are in fully compliance with DVD specs, so I guess that there aren't any complaint in use it.

[danpos]

Thanks. However can AQE work with the Qmatop off and produce the same result with it on as encoding with it on takes a very long time.

At the current stage, AQE is working very well either QMatOp ON or OFF. Regard on "same result" asked for you, the answer is: it will be so good as much you using yours 'feelings' and knowledge and so to choose the adequate custom quantum matrix for each part (Main Movie, Extras, etc). I already did it and got very good results. ;)

Greetings,

[masscamp24]

Thanks Danpos you have been a great help. I'm going to try it and let you know the result.

[danpos]

Thanks Danpos you have been a great help. I'm going to try it and let you know the result.

Your welcome! :)

Greetings,

[raquete]

wow,long threads and links to read about aqe,danpos.

searching topics with matrices inside aqe forum i found that thead where have one log file posted in rapidshare and i download it.is one long log that search for perfect size or something like this.the matrix found to be used scares me because have too much high values like 215,237,220.seems that aqe search for final size and adapt the better matrix for it and not for quality but i can be wrong,i need more lecture. :o

this is the thread of the new version with the log to download in the last posthttp://www.vmesquita.com/forum/index.php?topic=4836.15

thank you. :)

@ Sunhillow
one screenshot from source and another from result will be cool using vdub frame(i) with fire and moviments.

thank you. :)

[Sunhillow]

Raquete,
I will look for some screenshots, but cannot post them before monday evening CET

[danpos]

wow,long threads and links to read about aqe,danpos.

searching topics with matrices inside aqe forum i found that thead where have one log file posted in rapidshare and i download it.is one long log that search for perfect size or something like this.the matrix found to be used scares me because have too much high values like 215,237,220.seems that aqe search for final size and adapt the better matrix for it and not for quality but i can be wrong,i need more lecture. :o

this is the thread of the new version with the log to download in the last posthttp://www.vmesquita.com/forum/index.php?topic=4836.15

thank you. :)




This was a previous release and SAPSTAR fixed and improved the QMatOp engine, so no more appears the so high values that you saw from logs there ... ;)

Greetings,

[raquete]

the log file and the whole page 2 in the link was posted after the last aqe update...but it's ok danpos. thanks. :)

@ Sunhillow
what aqe version was used when you get that results?
do you still have the log file?i'm really curious to know how aqe work.
thank you. :)

[Sunhillow]

Hi Raquete,
this is an AQE logfile, although from another movie:
(I did cut some parts to make it shorter)

Width : 720 ; Height : 576 ; Bitrate Equivalent : 3244999
QMatOp feature enabled.
Normal Bitrate mode engaged
Pass 1/4
Bitrate : 3894000
D:\TMP\D2VAVS\V01001100001012.avs: 720x576, 25000/1000 fps, 858 frames
converting YUY2 -> YV12
Assigned new YUV clip
converting YUY2 -> RGB24
Assigned new RGB clip
AVS File opened successfully
Chosen multiplier for QMatop : 5
Creating : C:\Intra0.ppm
Input Uncompressed File Size: 1244175
(Bitrate Optimization) Target File Size: 38912
(Bitrate Optimization) File Size Tolerance: 1945 (5.0 Percent)
(Bitrate Optimization) Iteration 1 Target Perceptual Error: 1.0000000000
--Compressed JPEG File Size: 29611
(Bitrate Optimization) Iteration 2 Target Perceptual Error: 0.5000000000
--Compressed JPEG File Size: 40671
Target Perceptual Error: 0.5000000000
Compressed JPEG File Size is: 40671
Compression Ratio: 30.591207
Creating : C:\Inter1.ppm
Input Uncompressed File Size: 1244175
Target Perceptual Error: 0.5000000000
Compressed JPEG File Size is: 185094
Compression Ratio: 6.721855
...
Creating : C:\Intra60.ppm
Input Uncompressed File Size: 1244175
(Bitrate Optimization) Target File Size: 38912
(Bitrate Optimization) File Size Tolerance: 1945 (5.0 Percent)
(Bitrate Optimization) Iteration 1 Target Perceptual Error: 1.0000000000
--Compressed JPEG File Size: 50739
(Bitrate Optimization) Iteration 2 Target Perceptual Error: 50.5000000000
--Compressed JPEG File Size: 8045
(Bitrate Optimization) Iteration 3 Target Perceptual Error: 25.7500000000
--Compressed JPEG File Size: 8355
(Bitrate Optimization) Iteration 4 Target Perceptual Error: 13.3750000000
--Compressed JPEG File Size: 8780
(Bitrate Optimization) Iteration 5 Target Perceptual Error: 7.1875000000
--Compressed JPEG File Size: 12288
(Bitrate Optimization) Iteration 6 Target Perceptual Error: 4.0937500000
--Compressed JPEG File Size: 20518
(Bitrate Optimization) Iteration 7 Target Perceptual Error: 2.5468750000
--Compressed JPEG File Size: 28276
(Bitrate Optimization) Iteration 8 Target Perceptual Error: 1.7734375000
--Compressed JPEG File Size: 35452
(Bitrate Optimization) Iteration 9 Target Perceptual Error: 1.3867187500
--Compressed JPEG File Size: 41578
(Bitrate Optimization) Iteration 10 Target Perceptual Error: 1.5800781250
--Compressed JPEG File Size: 38038
Target Perceptual Error: 1.5800781250
Compressed JPEG File Size is: 38038
Compression Ratio: 32.708739
Creating : C:\Inter61.ppm
Input Uncompressed File Size: 1244175
Target Perceptual Error: 1.5800780000
Compressed JPEG File Size is: 128793
Compression Ratio: 9.660269
...
Creating : C:\Intra840.ppm
Input Uncompressed File Size: 1244175
(Bitrate Optimization) Target File Size: 38912
(Bitrate Optimization) File Size Tolerance: 1945 (5.0 Percent)
(Bitrate Optimization) Iteration 1 Target Perceptual Error: 1.0000000000
--Compressed JPEG File Size: 52812
(Bitrate Optimization) Iteration 2 Target Perceptual Error: 50.5000000000
--Compressed JPEG File Size: 8079
(Bitrate Optimization) Iteration 3 Target Perceptual Error: 25.7500000000
--Compressed JPEG File Size: 8439
(Bitrate Optimization) Iteration 4 Target Perceptual Error: 13.3750000000
--Compressed JPEG File Size: 8867
(Bitrate Optimization) Iteration 5 Target Perceptual Error: 7.1875000000
--Compressed JPEG File Size: 12133
(Bitrate Optimization) Iteration 6 Target Perceptual Error: 4.0937500000
--Compressed JPEG File Size: 20633
(Bitrate Optimization) Iteration 7 Target Perceptual Error: 2.5468750000
--Compressed JPEG File Size: 28966
(Bitrate Optimization) Iteration 8 Target Perceptual Error: 1.7734375000
--Compressed JPEG File Size: 37312
Target Perceptual Error: 1.7734375000
Compressed JPEG File Size is: 37312
Compression Ratio: 33.345170
Creating : C:\Inter841.ppm
Input Uncompressed File Size: 1244175
Target Perceptual Error: 1.7734380000
Compressed JPEG File Size is: 140581
Compression Ratio: 8.850236
Normalized matrix
8 9 9 10 14 55 165 191
8 10 9 10 13 37 163 191
8 9 13 17 25 141 191 191
10 10 17 96 178 191 191 191
15 14 87 191 191 191 191 191
28 142 191 191 191 191 192 225
175 191 191 191 191 192 225 255
191 191 191 191 191 225 255 255
16 16 16 17 19 21 24 29
16 17 17 17 18 19 21 25
16 17 19 19 20 21 24 28
17 17 19 22 24 26 29 34
19 18 20 24 28 33 39 50
21 19 21 26 33 46 68 97
24 22 24 29 41 68 122 149
30 25 28 35 60 105 150 164
Intra QuadriLinear Equation : -0.000391373X^4 + 0.0350191X^3 + -0.641686X^2 + 3.21765X + 8
Inter QuadriLinear Equation : 1.7255e-005X^4 + -0.000640763X^3 + 0.00764804X^2 + 0.0959899X + 16
Linearized matrix
8 9 12 10 8 10 108 123
12 10 9 10 13 94 138 191
12 9 13 17 81 155 191 191
10 10 21 69 172 191 191 191
15 26 59 191 191 191 191 191
33 49 209 191 191 191 192 225
40 228 191 191 191 192 225 255
248 191 191 191 191 225 255 255
16 16 16 17 19 17 20 21
16 17 17 17 18 19 21 39
16 17 19 19 20 21 37 42
17 17 19 19 24 35 45 83
19 18 20 24 33 48 77 89
18 19 24 31 51 72 95 134
18 25 29 55 67 102 125 143
27 28 58 63 109 117 153 164
Darkness level : 0.172866
Dark Limit : 0.0730125
Darkmasking value : 0
Quantizer value : 14 (delta: -3)
Qmin : 1 ; Qmax : 14
Tolerance : 1961333
Adjusted Min bitrate: 952001
Done
Normal Bitrate mode engaged
Pass 2/4
Bitrate : 3894000
D:\TMP\D2VAVS\V01001100001012.avs: 720x576, 25000/1000 fps, 858 frames
converting YUY2 -> YV12
Assigned new YUV clip
converting YUY2 -> RGB24
Assigned new RGB clip
AVS File opened successfully
8 9 12 10 8 10 108 123
12 10 9 10 13 94 138 191
12 9 13 17 81 155 191 191
10 10 21 69 172 191 191 191
15 26 59 191 191 191 191 191
33 49 209 191 191 191 192 225
40 228 191 191 191 192 225 255
248 191 191 191 191 225 255 255
16 16 16 17 19 17 20 21
16 17 17 17 18 19 21 39
16 17 19 19 20 21 37 42
17 17 19 19 24 35 45 83
19 18 20 24 33 48 77 89
18 19 24 31 51 72 95 134
18 25 29 55 67 102 125 143
27 28 58 63 109 117 153 164
Darkness level : 0.172866
Dark Limit : 0.0730125
Darkmasking value : 0
Quantizer value : 14 (delta: -3)
Qmin : 1 ; Qmax : 14
Tolerance : 1961333
Adjusted Min bitrate: 952001
Count Q1 : 453 , Count Q2 : 35
Calculated Min Qmin : 1 ; Max Qmax : 19
Done
Normal Bitrate mode engaged
Pass 3/4
Bitrate : 3894000
D:\TMP\D2VAVS\V01001100001012.avs: 720x576, 25000/1000 fps, 858 frames
converting YUY2 -> YV12
Assigned new YUV clip
converting YUY2 -> RGB24
Assigned new RGB clip
AVS File opened successfully
8 9 12 10 8 10 108 123
12 10 9 10 13 94 138 191
12 9 13 17 81 155 191 191
10 10 21 69 172 191 191 191
15 26 59 191 191 191 191 191
33 49 209 191 191 191 192 225
40 228 191 191 191 192 225 255
248 191 191 191 191 225 255 255
16 16 16 17 19 17 20 21
16 17 17 17 18 19 21 39
16 17 19 19 20 21 37 42
17 17 19 19 24 35 45 83
19 18 20 24 33 48 77 89
18 19 24 31 51 72 95 134
18 25 29 55 67 102 125 143
27 28 58 63 109 117 153 164
Darkness level : 0.172866
Dark Limit : 0.0730125
Darkmasking value : 0
Quantizer value : 14 (delta: -3)
Qmin : 1 ; Qmax : 14
Tolerance : 1961333
Adjusted Min bitrate: 952001
Count Q1 : 453 , Count Q2 : 35
Calculated Min Qmin : 1 ; Max Qmax : 19
Expected size : 16705260 ; Real Size : 18499315 ; Error : -11 %
Quality Factor correction : 0.90302
Corrected Qmin : 1 ; Corrected Qmax : 23
New Tolerance : 1961333
Done
Normal Bitrate mode engaged
Pass 4/4
Bitrate : 3894000
D:\TMP\D2VAVS\V01001100001012.avs: 720x576, 25000/1000 fps, 858 frames
converting YUY2 -> YV12
Assigned new YUV clip
converting YUY2 -> RGB24
Assigned new RGB clip
AVS File opened successfully
8 9 12 10 8 10 108 123
12 10 9 10 13 94 138 191
12 9 13 17 81 155 191 191
10 10 21 69 172 191 191 191
15 26 59 191 191 191 191 191
33 49 209 191 191 191 192 225
40 228 191 191 191 192 225 255
248 191 191 191 191 225 255 255
16 16 16 17 19 17 20 21
16 17 17 17 18 19 21 39
16 17 19 19 20 21 37 42
17 17 19 19 24 35 45 83
19 18 20 24 33 48 77 89
18 19 24 31 51 72 95 134
18 25 29 55 67 102 125 143
27 28 58 63 109 117 153 164
Darkness level : 0.172866
Dark Limit : 0.0730125
Darkmasking value : 0
Quantizer value : 14 (delta: -3)
Qmin : 1 ; Qmax : 14
Tolerance : 1961333
Adjusted Min bitrate: 952001
Count Q1 : 453 , Count Q2 : 35
Calculated Min Qmin : 1 ; Max Qmax : 19
Expected size : 16705260 ; Real Size : 18170729 ; Error : -8.8 %
Quality Factor correction : 0.830192
Corrected Qmin : 2 ; Corrected Qmax : 27
New Tolerance : 1961333
Done
Final Result :
Expected size : 16705260 ; Real Size : 16622034 ; Error : 0.5 %

The DVDRB logfile of Episode 3 is this:
- Detected ANGLE and/or INTERLEAVING on source.
-----------------
[12:33:54] Phase I, PREPARATION started.
- CCE SP 0.31.1.1 encoder selected.
- "CCE Adaptive Quantizer Matrices" is enabled.
- "Steal Space from Extras" mode is enabled.
- "Movie Only" mode is enabled.
- VTS_04: 3.549.483 sectors.
-- ANGLE and/or INTERLEAVING is present.
-- Scanning and writing .D2V & .AVS files
- Processed 132,1MB ILVU section...
-- Processed 203.699 frames.
-- Building .AVS and .ECL files
- Reduction Level for DVD-5: 59,8%
- Overall Bitrate : 3.587Kbs
- Space for Video : 3.567.316KB
- HIGH/LOW/TYPICAL Bitrates: 4.510/600/3.587 Kbs
[12:48:30] Phase I, PREPARATION completed in 15 minutes.
[13:04:35] Phase II ENCODING started
- Creating M2V for VTS_04 segment 0
- Creating M2V for VTS_04 segment 1
[13:06:06] Phase II ENCODING started
- Creating M2V for VTS_04 segment 0
- Creating M2V for VTS_04 segment 1
- Creating M2V for VTS_04 segment 2
.....
- Creating M2V for VTS_04 segment 71
- Creating M2V for VTS_04 segment 72
- Creating M2V for VTS_04 segment 73
- Creating M2V for VTS_04 segment 74
- Extracting STILLS for VTS_04 segment 75
[19:34:25] Phase II ENCODING completed in 388 minutes.
[08:49:22] Phase III, REBUILD started.
- Processing VTS_04
- Reading/processing TMAP table...
- Rebuilding seg 0 VOBID 1 CELLID 1
- Rebuilding seg 1 VOBID 1 CELLID 1
- Processed 132,1MB ILVU section...
- Updating NAVPACKS for VOBID_01
- Rebuilding seg 2 VOBID 2 CELLID 1
- Rebuilding seg 3 VOBID 2 CELLID 2
- Rebuilding seg 4 VOBID 3 CELLID 1
- Rebuilding seg 5 VOBID 3 CELLID 2
- Updating NAVPACKS for VOBID_02
- Updating NAVPACKS for VOBID_03
- Rebuilding seg 6 VOBID 4 CELLID 1
- Rebuilding seg 7 VOBID 4 CELLID 2
- Rebuilding seg 8 VOBID 4 CELLID 3
- Rebuilding seg 9 VOBID 4 CELLID 4
- Rebuilding seg 10 VOBID 4 CELLID 5
- Rebuilding seg 11 VOBID 4 CELLID 6
...
- Rebuilding seg 34 VOBID 4 CELLID 29
- Rebuilding seg 35 VOBID 4 CELLID 30
- Rebuilding seg 36 VOBID 4 CELLID 31
- Updating NAVPACKS for VOBID_04
- Rebuilding seg 37 VOBID 5 CELLID 1
- Rebuilding seg 38 VOBID 5 CELLID 2
...
- Rebuilding seg 73 VOBID 5 CELLID 37
- Rebuilding seg 74 VOBID 5 CELLID 38
- Updating NAVPACKS for VOBID_05
- Rebuilding seg 75 VOBID 6 CELLID 1
- Updating NAVPACKS for VOBID_06
- WARNING: VOB Count Old: 0 New: 6
- Updated VTS_C_ADT.
- Updated VTS_VOBU_ADMAP.
- Updated IFO: VTS_01_0.IFO
- Correcting VTS Sectors...
[09:14:58] Phase III, REBUILD completed in 25 minutes.

Done.
-----------------

[Sunhillow]

So here are 2 screenshots with high motion and explosion :p

Original, time 00:02:47
http://img60.imageshack.us/img60/9113/000247ori7ya.th.png (http://img60.imageshack.us/my.php?image=000247ori7ya.png)

AQE, time 00:02:47
http://img166.imageshack.us/img166/5246/000247aqe5al.th.png (http://img166.imageshack.us/my.php?image=000247aqe5al.png)

Original, time 01:02:54
http://img174.imageshack.us/img174/1250/010254ori5rx.th.png (http://img174.imageshack.us/my.php?image=010254ori5rx.png)

AQE, time 01:02:54
http://img174.imageshack.us/img174/5663/010254aqe3kx.th.png (http://img174.imageshack.us/my.php?image=010254aqe3kx.png)

The first screenshot shows strong blocks, the second looks a bit more granular than the original.
Maybe the use of a very light smooth filter would have helped.
Again, the reduction level was 59,8%

[Carpo]

One disadvantage of this encoder is its slowness, but the quality was really good for a compression of ~63% IIRC.

i have been testing it since it first arrived (well havent tested it much lately due to illness i have had for pass 4 months) but i can say after grabbing the new beta that it is conciderably faster than what it used to be (encoding part anyway), used to moan at sapstar about it :cool: - but in the end its down to what you want.

If you want two passes and quality then i suggest you use HC (or dare i say it a one click solution which wont give you quality but will give you speed)

If you want top quality encodes at the cost of speed then go for AQE or CCE

I use (depends on my mood at the time) cce or aqe with iDCT=3 (nans dg dll) with avamat6 matrice for main (3500-9000) and avamat7 (1800-3500) for other areas (dont really leave the extras on the disc unless they are something special) and often have CCE Adaptive Quantizer Matrices on as default after being advised that it does no real harm to the film (just some players dont respect dvd spec and wont play).

I have used this for a fair few backups 250+ and have always had what i concider good results (always look good on the tv - lcd pc monitor tends to be picky)

But the settings you use have to be what you concider to be good for you - these are merely guidelines but be warned if you chose iDCT option 3 or 4 they will slow your encode down a far bit - but will give you better results (more often than not)

[raquete]

:) hi Sunhillow
good pictures,a pity that you get B frame and not from I frame.
the result is not bad(granular) as you posted
i did screenshots from my dvd source and they have better definitions comparing with your source and result.did you resized the pictures before host?if you did you lose quality.

i did(for test only)with dvdshrink full disk(custom radio in menus/extras,automatic compression in main movie(50.3%)deep analizes and sharp default)removing spanish subs/audio track.the result is better comparing screenshots with your source and result.

screenshots from vdubmod to paint from frame 8436(b),just the same as yours but my dvd is ntsc(720*480):

source
http://img117.imageshack.us/img117/4438/source4ha.th.png (http://img117.imageshack.us/my.php?image=source4ha.png)

dvdshrink
http://img509.imageshack.us/img509/6728/shrink6yc.th.png (http://img509.imageshack.us/my.php?image=shrink6yc.png)

of course,encoders always give better result than dvdshrink at 50.3%.

aqe deserve better disclosure but that matrix with loud values in the log still scare me.

thank you.

[DK]

transcoding sw3 with shrink i was faced with quality that was far from tolerable

tons of artefacts even with aec max smooth and an average bitrate of ~3100kbps

[raquete]

transcoding sw2 with shrink i was faced with quality that was far from tolerable

tons of artefcts even with aec max smooth and an average bitrate of ~3100kbpsyou lose some details DK

...(for test only)with dvdshrink ....sharp default...
...the result is better comparing screenshots with your source and result.

of course,encoders always give better result than dvdshrink at 50.3%.now please,compare the results from aqe and dvdshrink.

[DK]

now please,compare the results from aqe and dvdshrink

me?

i did and i know that shrinks results looked crappy (better tsill than all other transcoderts) when i tested sw3 with a couple of transcoders (dvd2one v2, shrink, clonedvd + rejig) vs rebuilder using procoder

[raquete]

did you read the whole thread? :rolleyes: and my last post?
the pictures are here to compare.
don't need to encode or shrink,just click in each pictures posted.

[Sunhillow]

H Raquete,

I made the screenshots with MP Classic, as VirtualDub did not load the DVD stream. Will try using VDubMod soon :) to get I frames


aqe deserve better disclosure but that matrix with loud values in the log still scare me.

The matrices are from Fellowship of the Ring, extended edition, both discs to 1 DVD5. Here the bitrates are HIGH/LOW/TYPICAL: 3.994/1.303/2.559 Kbs

Just want to know how this will look :D

[DK]

i did read the whole thread, raquete, i saw your screenshot and i know what my comparisons looked like

[raquete]

i did read the whole thread, raquete, i saw your screenshot and i know what my comparisons looked like
DK,seems that Sunhillow resize the pictures before or when was hosting and lose quality.
look the size of Sunhillow"s screenshot(aqe 461k) and my screenshots(1013k)
is impossible that aqe picture have better look for your taste

do you know what i mean now? :p

[Sunhillow]

DK,seems that Sunhillow resize the pictures before or when was hosting and lose quality.

No, did not resize the images. PAL source is 720 x 576, which looks a bit squeezed without resizing. But as mentioned, I saved them from Mediaplayer Classic. Maybe this reduced the quality.

The images were converted from BMP to PNG using IrfanView, maybe this explains the size difference.

Here are 2 I-Frames:

Original
http://img511.imageshack.us/img511/6476/iframeori1vz.th.png (http://img511.imageshack.us/my.php?image=iframeori1vz.png)

encoded
http://img511.imageshack.us/img511/9453/iframeaqe6md.th.png (http://img511.imageshack.us/my.php?image=iframeaqe6md.png)

[DK]

raquete, i wasnt saying that anybodys picture looked better or worse - i said MINE do

i surely wont compare sunhillow's pictures with yours as i i wouldnt know who of you got rid of which audiostreams or what settings have been used on the menu and so on

therefor i prefer to compare for myself

[raquete]

No, did not resize the images. PAL source is 720 x 576, which looks a bit squeezed without resizing. But as mentioned, I saved them from Mediaplayer Classic. Maybe this reduced the quality.

The images were converted from BMP to PNG using IrfanView, maybe this explains the size difference.
seems that nobody read with atention.
i posted that my source is ntsc and i knew that your pictures came from pal
but my dvd is ntsc(720*480)
i'm not talking about how it looks squeezed in the screen but my pictures have 1013kb each and your's not. when you converted from bmp to png you loose quality!
you have to host bmps without convertion because imageshack will convert it to png.

DK,
my settings are posted full disk(custom radio in menus/extras,automatic compression in main movie(50.3%)deep analizes and sharp default)removing spanish subs/audio track.

:)

[Sunhillow]

you have to host bmps without convertion because imageshack will convert it to png

I converted a bmp to png and back - it was binary identical to the primary bmp.

[raquete]

then,if they have binary identical and are showing the perfect appearance, your results are worse than mine.
my screenshots have superior quality.

;)

[SAPSTAR]

For your info...I just released in the beta testers area a new version 0.32.0.1....This version seems to be very promising, up to now I had very good results. I think it worth a try...The visual quality should be improved...I improved the QMatOp engine....

[rendez2k]

Thanks Sapstar - when I get a chance I'll re-run my Gladiator encode to see if I can improve the opening credits as they looked a little blocky on a previous test (a difficult one I guess with a fire type effect). I take it you recommend to always use the QMatOp function and with 4 passes (its just it takes so long!)?

[SAPSTAR]

Thanks Sapstar - when I get a chance I'll re-run my Gladiator encode to see if I can improve the opening credits as they looked a little blocky on a previous test (a difficult one I guess with a fire type effect). I take it you recommend to always use the QMatOp function and with 4 passes (its just it takes so long!)?
With the last alpha version, only 2 passes are needed most of the time...and the QMatOp part is faster now.

[Sunhillow]

Hi Jérôme,
thanks for your efforts - just started a new encode of Episode 3 with the same streams left (main movie only, video 2 angles, audio german and english, subtitles german and english).

One thing attracted my attention when the transcoding started: for the opening with the scrolling text used a matrix with only "8" as values and was greatly oversized by > 50% in the second pass. Every pass brought it closer to the expected bitrate, but I think the final result still was a bit too large. Probably this is ok as it only affects a very small part of the movie.

Cheers!

[SAPSTAR]

Hi Jérôme,
thanks for your efforts - just started a new encode of Episode 3 with the same streams left (main movie only, video 2 angles, audio german and english, subtitles german and english).

One thing attracted my attention when the transcoding started: for the opening with the scrolling text used a matrix with only "8" as values and was greatly oversized by > 50% in the second pass. Every pass brought it closer to the expected bitrate, but I think the final result still was a bit too large. Probably this is ok as it only affects a very small part of the movie.

Cheers!
For small segments dont pay attention to the percentage, I show it only for info, explanation :
A 5% error on a 2MB segment is 50KB as you can see its negligible on a whole DVD. Moreover, on any very small segment it's harder to keep a good rate as the number of frames limits the variations. So dont worry the movie should end OK. But you should pay attention to the big segments, there a slight percentage can be a big size.
Thank you for your support, and I look forward to having your results !!

[EDIT] The matrices are normal :) The new ones are fully MPEG2 compliant but are strictly using the specs not less like many matrices... (nothing says that the inter matrices must be >16)

[EDIT2] Let me know ablut the final visual quality, I had great results on my own tests....

[Carpo]

For your info...I just released in the beta testers area a new version 0.32.0.1....This version seems to be very promising, up to now I had very good results. I think it worth a try...The visual quality should be improved...I improved the QMatOp engine....

u could have released that a day earlier could have done my test encode with it ;)

[Sunhillow]

Here is a new screenshot made with SAPSTAR's version 0.32.0.1.

http://img504.imageshack.us/img504/6255/aqenew4fi.th.png (http://img504.imageshack.us/my.php?image=aqenew4fi.png)

and the old one for comparison:
http://img504.imageshack.us/img504/6505/aqeold1yn.th.png (http://img504.imageshack.us/my.php?image=aqeold1yn.png)

I chose the same B-frame as with the first screenshots at approx. 02:47 in the 25 FPS PAL version - not only I-frames should look good :)
The screenshots were made with VirtualDub MPEG2, conversion to PNG with IrfanView.

The new encoder indeed causes less blocking, the screenshot looks smoother - but not blurred :)

The total size of the DVD files is 4.667.805.696 Bytes, with AQE 0.31 it was
4.628.289.536 Bytes.

[raquete]

hi Sunhillow,
good to see you back.
:)

about your screenshots,i have one idea.

your pictures was converted from bmp to png and when they was hosted,maybe imageshack convert it again and you loose details.
this is what i mean in my old post here.your pictures have 445 and 469kb and my pictures have 1013kb and this is the reason why mine are showhing better quality.
all i want to tell you that your screenshots are not so good and not about the result of aqe.
if you agree,i want to host few seconds of the movie.(video only demuxed/DGIndex)
you download the file,encode and host again.do the same with your source because my source is ntsc.doing that i will see the real result and not from screenshots.

can you do that?
:)

[Sunhillow]

Ok, Raquete,

let's try it :)

Today I wanted to upload the screenshots as bmp, but they were too big (~1.3 MByte)

Edit: Is there a way to save a single frame with VirtualDub? I used "Copy source frame to clipboard"

Edit2: My coworker loves DVD2One. Here is the same frame from his transcode:
http://img452.imageshack.us/img452/2125/dvd2one4rn.th.png (http://img452.imageshack.us/my.php?image=dvd2one4rn.png)

He left only the german audio and subtitles, so compression was ~70%

[rendez2k]

Edit2: My coworker loves DVD2One. Here is the same frame from his transcode


Erm, that shot looks awful!
I'm just testing a 4 pass encode with the new Doctor Who boxset... will let you know the results!

[raquete]

rendez2k,
this picture from dvd2one is horrible! :p
all sparks are twisted,solarizantions in the fire....too bad.

Sunhillow
i knew that your pictures was too big for host in imageshack,it's ok.

to save a single frame in vdubmod or vdubmpeg2 is simple.
get 2 pictures,one from the dvd ripped and another from your encode.
load the first vob of the main movie and select one audio track .(ignore all advices from vdub)
edit,go to...,write 8436 in jump the frame number.
that is the picture but is (b) frame and better is change to (I) frame.
click in the the arrow with a key[key previous]and you get the (I) frame.
Frame 8431 (00:05:51.643) (I)
now in the video menu,copy source frame to the clipboard (Ctrl-1).
open paint,paste the picture and save as .bmp.
do the same with your encoded file from aqe.
create one zip or rar file with the pictures,host in rapidshare and post the link here.
this is faster than cutting vobs and encoding.
i will do the same with my sources.

thank you.

[Sunhillow]

raquete,

could you post a screenshot of the position you are talking about? In the PAL version frame 8431 is no I-frame, and position 00:05:51.643 is not frame 8431 :scared:

... edit: well, some posts before you wrote frame no. 8436 is the frame I posted. In the PAL version it is no. 6315 :p

in the DVDRB encode 6313 is the next I frame, in the original DVD source it is 6316.

:rolleyes:

[raquete]

:eek: i knew and forgot this detail,sorry.

in SWIII the positions are differents in ntsc and pal dvds(don't know if happen the with all movies)
to find the right position,load the (old)frame where you got the screenshots and click in the the arrow with a key[key previous]and you get the right (I) frame.
is the first (I) frame before the old frame posted here,ok?
(have action and fire,cool for tests.)
:)

[Sunhillow]

in SWIII the positions are differents in ntsc and pal dvds(don't know if happen the with all movies)

The positions can never be the same. PAL plays the 24 fps FILM a bit too fast (at 25 fps), for NTSC some queer frame repetition rule is used to get 60 subframes which are then combined to 30 interlaced frames.

to find the right position,load the (old)frame where you got the screenshots and click in the the arrow with a key[key previous]and you get the right (I) frame.

That is what I did

[Sunhillow]

Neither Imageshack nor IrfanView do not alter my images. I made a screenshot, saved it as BMP using M$Paint, did the same with IrfanView and saved as PNG. Then converted the PNG to BMP, again with IrfanView and did a binary compare - no differences.
Then uploaded PNG to Imageshack, downloaded it, saved again as BMP and compared. You guess? No difference.

raquete, here is I-frame no. 6313 from the transcode:
http://img476.imageshack.us/img476/8183/E3_6313_AQE.th.png (http://img476.imageshack.us/my.php?image=E3_6313_AQE.png)

and the corresponding B-frame no. 6313 from the original:
http://img476.imageshack.us/img476/3339/E3_6313_ORI.th.png (http://img476.imageshack.us/my.php?image=E3_6313_ORI.png)