New Quantization Matrices - EQM V3 series (last update: 04/02/2005 - V3ULR)

[Heini011]

Hi Beave,

denoise the hd clip (for much noise use mvdenoise from mvtools) and use eqmv3hr instead of 6o9. (but eqmv3hr is soften the clip a little bit.)

greetings.

[Sharktooth]

Maybe you want to use EQM V2 (autogk sharp matrix) for that purpose since it scales very well... but if it's "too much", try ULR (theoretically it shoud do it even better). However ensure you denoise the source without loosing details (read: use the right filters and not just the first denoiser you find on the web).

[Soulhunter]

@ Beave

If you dont mind low fps, give FFT3D a try!


Bye

[Sharktooth]

The reason is HD eats up a lot of bitrate and having some extra compressibility by denoising would ensure a much better encode without wasting an enormous amount of bits... :)

[Didée]

True. However my fear is, there are people out there that brute-filter so much compressibility on their HD sources that, in the end, they're not encoding HD anymore, but just big framesizes ...

[Soulhunter]

Yeah, thats the reason why i suggested FFT3D for the denoising... ^^

It kills most of the noise without removing too much of the "real" details !!!


Bye

[BigDid]

Hello all,

Not enough practice to be affirmative, but I am actually testing fft3d compared to peachsmoother.
fft3d default (sigma=2) is comparable to peach default but really retains more details.
For more compression, try sigma=3.

Sizes of one of my testing (original mpg2 being 110672ko, encoded to 75% xvid, one pass):
Avi no filter : 30890ko, 13,83fps
Avi peach() : 28456ko, 13,14fps
Avi fft3d s=2: 28378ko, 8,38fps
Avi fft3d s=3: 26764ko, 8,44fps

There is also a quicker gpu version of fft3d for those having an ati radeon 95xx or higher or geforce 5xxx at least or any gpu having hardware supporting for dx90c.

Did

[Beave]

Thanks, I usually only put removegrain() in the avs. I just got a few eps with extrem grain in it. I only get 25% compressibility without denoisers.
I tried lremovedust, but only lremovedust(17,10) got rid of most of the noise and got compressibility to 60%. But now the clip looks pretty washed out. I will try your suggested filters. Maybe using the right matrix I could get away with 45%.

But usually the hd stuff isn't as bad so removing some of the filmgrain should help the picture quality. I just wondered if there is a matrix, that is especially good for noisy and blocky sources.

So FFT3D seems to be the latest in denoising? It would be great to remove some of the noise without loosing detail, which is the main reason for HDTV anyways.

[Soulhunter]

FFT3D gives excellent results as long you dont "overdo" it... For most DVD sources simgma=1 should be already enough, but if you cope with very grainy/noisy stuff you'd need much higher sigma values to kill the noise... Well, the catch is that FFT3D produces slight grid artifacts if you use too high sigma values! So, using FFT3D with sigma=2 + some mild tempo-spatial denoising afterwards could be better than using FFT3D with sigma=3 alone for example... ^^


Bye

[Backflip]

Let's say you have a 2.35:1 AR movie to compress.
Cropping the black bars (from the original 720x576 res) results in a 720x432 picture with an incorrect aspect ratio.
Generally to get it right the 432 is resized to 308 (nearest mod16 value) so you obtain a 720*308 picture that looks with the correct AR.
Doing this you're loosing pixels in favor of a correct looking picture.
There's another way (anamorphic encode) to get the correct AR without loosing pixels and almost all DVDs movies use this technique to NOT loose definition.

Edit - cleaned up :)

I believe I get the way to correctly encode anamorphically while keeping all picture info and AR -- Set AR in the XviD codec (240,100 (2.40:1) for example), crop out black bars (only down to 720 x 432) giving 0% AR error. Set Input Pixel PAR to 1:1.

I encoded The Pacifier DVD with:
2-pass, EQM V3ULR rev 3, QPEL, B-VOPs - 2/1.5/1, Packed Bitstream, Closed GOV, Chroma Motion, Motion Search Precision (Ultra High), VHQ Mode (4) Chroma Optimizer and Trellis Quantization (1,31), Lanczos

It had a compressibility of 87% at 720 x 432 anamorphic for a 1404 MB target size. High, but not too high I guess.

To me it came out better than the 1404 MB/MPEG matrix encode. Particularly the first boat/water scene (which is probably the worst scene to encode). It's not as blocky.

Ended up with a 1.37 GB (1,475,354,624 bytes) AVI and muxed into MKV @ 702MB + 701MB with MMG.

Used same settings for Assault On Precinct 13 DVD except this movie was less compressible @ 77%. Got an ever so slightly oversized file 1.37 GB (1,475,850,240 bytes) AVI which I'm not entirely sure about. MMG will only output 703 + 700 MB (with settings at 702 MB (2 file max.)

Anyways I'm pleased with the results, thank you :)

[Sharktooth]

:) thanks

[Backflip]

Bit off-topic, but I've used 720 x 432 for a 2.35:1 and 2.40:1 DVD (is it right to use 720 x 432 for 2.40:1 DVD?), I've now got a 1.85:1 DVD which I'm ripping (The Ring Two), what would be the closest MOD 16 values? 720 x 560 seems to be it (AR error is 0). Is this correct? I have cropping of 8 and 8, top & bottom. Top border is quite large

Are the matrices good for 720 x 560? I think they are, but I'm asking to be sure, thanks :)

[Sharktooth]

well, you just really need to crop the black borders to a mod 16 resolution and set the AR in the codec - ONLY -. no resize, no fancy things... the final AR will be correct.
of course anamorphic resolution encodes have more pixel than a non anamorphic one, so remember to choose a slightly lower bitrate matrix.

[Backflip]

Yip, I'm using EQM V3ULR. The compressibility of the matrix is very high. I'm getting 85.8% of 0.171 and that's with QPEL off (other options mostly defaults + VHQ 4/MSP 6), which seems reasonable. Bitrate is around 1400. MPEG yields 71.8%. Maybe numbers are a little low for full 720 x 460, but it seems reasonable. I got fairly good results with action movies at the same bitrate and compressibility (although I used QPEL in those cases), looked great on tv :) I need just to cut down on some things so it's playable (no dropping of frames and less early/late) on the Xbox via XBMC (FFMPEG decoder), so I took out QPEL.

If I go to like regular 720 x 384 the percentage is 125.1. Is there any problem with encoding with the percentage that high? What are the consequences?

Anyways, thanks for the matrices. Maybe I go to 3-4CD XviD/AAC backups giving about 1800-2400 kbps for full anamorphic :)

[Sharktooth]

If I go to like regular 720 x 384 the percentage is 125.1. Is there any problem with encoding with the percentage that high? What are the consequences?
It's ok but you will obtain an undersized file. It's the case to use a higher bitrate matrix.

Anyways, thanks for the matrices. Maybe I go to 3-4CD XviD/AAC backups giving about 1800-2400 kbps for full anamorphic :)
...using a higher bitrate matrix :)

[Backflip]

It's ok but you will obtain an undersized file. It's the case to use a higher bitrate matrix.

...using a higher bitrate matrix :)

Yip, I'm going for your ULR matrix now :), it's yielding around 93% with The Ring Two encoding. I used the MPEG matrix before, but that still seemed too blocky. Gonna be encoding at around 1695 kbps (XviD/AAC 6CH). I've been twisting and turning on whether to keep original AC3 or use 6CH AAC. I mean if the setting of Normal HE-AAC is infact percievably the same as the AC3 file (anyone have any benchmarks?), then why not :)

edit - I has founded the problemo. The Xbox just can't handle full (DVD res - 720 x 560) anamorphic XviD encodes @ beyond around 1600 kbps (even with QPEL and other CPU usage increasing things disabled). Woers me :(

Again though what parts I was able to watch certainly are less blocky than the MPEG matrix, so with the ULR matrix blockiness is majorly reduced. The MPEG matrix seems totally outclassed.

[Sharktooth]

Yip, I'm going for your ULR matrix now :), it's yielding around 93% with The Ring Two encoding. I used the MPEG matrix before, but that still seemed too blocky. Gonna be encoding at around 1695 kbps (XviD/AAC 6CH). I've been twisting and turning on whether to keep original AC3 or use 6CH AAC. I mean if the setting of Normal HE-AAC is infact percievably the same as the AC3 file (anyone have any benchmarks?), then why not :)
AAC is superior to AC3, but re-encoding leads always to a loss in quality. Also AAC is more complex to decode (slightly).

edit - I has founded the problemo. The Xbox just can't handle full (DVD res - 720 x 560) anamorphic XviD encodes @ beyond around 1600 kbps (even with QPEL and other CPU usage increasing things disabled). Woers me :(
Disable post-processing and use the latest XBMC.

Again though what parts I was able to watch certainly are less blocky than the MPEG matrix, so with the ULR matrix blockiness is majorly reduced. The MPEG matrix seems totally outclassed.
EQM V3ULR was not ever meant to replace MPEG matrix, however it obviously behaves much better than it at low bitrates.

[UltimAtomMAX]

I've updated my post about some odd xvid codec results and now reply here to say that custom matrix V3ULR worx on KISS DP-450 SA player.

[Backflip]

I've since found the real problem (from what I've read at the XBMC forum) with the 720 x 560 anamorphic encode I did. FFMPEG has difficulty with B-VOP's of 2. I've even seen this affect some lower resolution encodes (though they don't drop frames). With post-processing disabled the problem still occurs. Using the latest T3CH CVS release from a few days ago.

[Backflip]

The rips done using EQM V3 series are splendid. However I've noticed that straight black at certain points (i.e. the end or beginning credits) have some sort of slightly lighter shade of black running through them - the light black appears as the credits/words roll up. The lighter black is much less noticeable during the day, but at night I can see it pretty clearly; using XBMC, displaying on a 25" 4:3 TV.

I can't see any light black when using MPlayer and watching it on the PC monitor, but the monitor is smaller at 17" 4:3.

It seems to happen no matter what settings I use. Latest rip/encode was Sahara. I used:

Square Pixel (704 x 288 mod 16)
Adaptive Quant
QPel
GMC
B-VOP 2/1.5/1.0
Packed Bitstream
Closed GOV
eqm_v3ulr_rev31CD
Chroma Optimizer (BVOP sensitivity = 0)
Trellis (1,31)
MSP: 6
VHQ: 4 (I encode overnight anyway :))
Chroma Motion

Compressibility was 84%

I think I'm going to do another encode tonight so I'll use one of the default matrices and similar settings to the above^. Will report back :)

[Sharktooth]

can you post a screenshot or a small clip?

[Backflip]

I'm getting some screenshots ready and a clip of the credits. To be honest when I'm playing the files through XBMC is the only time when I see the lighter black, maybe it's to do with the default constrast settings, or just the build of MPlayer-XBMC is using? So it could be just an issue with XBMC decoding.

After giving it more serious thought XBMC is said to not be perfect at playing back XviD or DivX, so this might be a small consequence of that.

[Sharktooth]

Maybe.

[Backflip]

Hiya Sharktooth. I've been reading through the forums mostly about resize/filtering (Bicubic/Bilinear/Lanczos) and noticed you mentioned a preference for not resizing. How do you do this? I only see that you've have to use a resize filter in Gordian Knot (like Bicubic/Bilinear/Lanczos etc) Or do you mean something else?

Just an example of how I'm approaching things. Just to make sure I'm using the matrices in the way they're supposed to be used. Figures returned (compressing Bride and the Prejudice DVD). My aim is 2CD DVD backup, keeping as much resolution as resonably possible:

I crop getting 70 from top and bottom -- leaving quite a big black bar on left and very minor black line on top and bottom giving lowest AR error (no Smart Crop All at this point). Figures are:

Using V3ULR
Target size = 1400 MB

XviD settings:
XviD - 2-pass
Chroma Optimizer
Motion search precision -- 6
VHQ -- 4
Trellis (1,31)

720 x 304 (-0.8 aspect error)
Bit rate: 1384 Kbps
0.253 B/(P*F)
65.9% of 0.384
Soft Bicubic resize

What would you do next? I would personally add 1 bframe, so I can use that resolution and have a compressibility around 70% or perhaps lower resolution to 640 x 272 (no Smart Crop All) and use VHR (with possibly Lanczos). I would personally like to make the rip look good at a high res, but what would you recommend?

[Sharktooth]

The full resolution for a DVD is 720x576. Depending on the movie AR, black bars are added.
When i say "no resizing" i mean anamorphic encode: keeping the full DVD resolution removing the black bars only then set the proper aspect ratio in the codec options.
You will end up with resolutions like 720x432 but the final aspect ratio will be corrected by the decoder. No lines are lost during the resizing process.
A quick forum search (start with "anamorphic encoding") will highlight some threads with more info on the subject.
Now you can try to encode your Bride and the prejudice DVD with this technique .
Since there is more data to compress (due to the higher horizontal resolution) try using 1 or 2 bframes and turning on adaptive quantization.
Just a note: some codec features effects (like b-frames, AQ/psy, etc) that will affect quality in selected areas or frames to keep higher details on significative counterparts will be less visually "annoying" and more efficient on higher res encodes.

[Backflip]

When you say remove the black borders only do you mean cropping all of it away until you can't crop any more without cropping out good picture/video? Or crop down until the AR is 0% even when there's quite a bit of black bar left?

edit - ok I've cropped down the top and bottom to 720 x 432. Incidently I was certain a bit of the top is cropped out with the cropping autocrop gives (72 top and bottom).

Figures I'm getting are 0.178, 73.6 and 0.242. Shall see how it turns out, thanks :)

edit2 - finally, I think I get it. You keep settings how they, then just down the resolution, without using Smart Crop All. So you keep all of the picture without taking out good picture :)

[Cyberace]

FYI, it now sounds as if the FFmpeg developers have found the root cause why the visual quality is poor on some XviD encoder files when use FFmpeg/libavcodec to decode it instead of using the native XviD codec, see: http://mplayerhq.hu/piperma....ml#3641 (http://mplayerhq.hu/pipermail/ffmpeg-devel/2005-September/thread.html#3641) "[Ffmpeg-devel] libavcodec: possible QPEL decoding bug (Marius the Mad)", basically it seems like there is a bug in libavcodec QPEL decoding (or rather lack of QPEL decoding) which can be seen as poor quality or artifacts in players that uses FFmpeg to decode XviD that been encoded with QPEL, (like XBMC, and also in FFdshow for Windows PC's).

frame 154 from 311564.avi, decoded by
libavcodec: http://img379.imageshack.us/my.php?image=154a2pp.png
xvid: http://img380.imageshack.us/my.php?image=154x2az.png

frame 283 from 311564.avi (I can't see any difference on this one), decoded by:
libavcodec: http://img380.imageshack.us/my.php?image=283a1ve.png
xvid: http://img380.imageshack.us/my.php?image=283x3qj.png

frame 597 from 311564.avi, decoded by:
libavcodec: http://img380.imageshack.us/my.php?image=597a1bc.png
xvid: http://img380.imageshack.us/my.php?image=597x2gz.png

Frame 154 from 310058.avi sample, decoded by:
libavcodec: http://img284.imageshack.us/img284/6352/n154a7bm.png
xvid: http://img284.imageshack.us/img284/4498/n154x2mj.png

Frame 283 from 310058.avi sample, decoded by:
libavcodec: http://img284.imageshack.us/img284/9566/n283a5ry.png
xvid: http://img140.imageshack.us/img140/5461/n283x1dg.png

Frame 597 from 310058.avi sample, decoded by:
libavcodec: http://img140.imageshack.us/img140/1294/n597a9id.png
xvid: http://img140.imageshack.us/img140/2001/n597x0yf.png

I didn't use any postprocessing (apart from YV12->RGB32
convertion during decoding).

Try to check it in a darkened room, possibly at night, the problem is very visible then. Try increasing the gamma value. If you use kuickshow as image viewer you could press "g" four times to lighten the images up. Then you will see that either libavcodec or xvid is faulty. btw. you can use gimp to show the difference between both images. The difference image won't be all black. Increase the gamma value to brighten the differences.


My clips triggering this behaviour is available here (below 10 MB):
http://uploadhut.com/view.php/311564.avi (1000 frame clip encoded with only only QPEL enabled)
http://uploadhut.com/view.php/310058.avi (encoded with qpel, gmc and mpeg quantizers)
The effect is extremely visible during the playback (much more so than in the screenshots).
So XviD files that haven't been encoded QPEL don't have this very noticable image quality (noisy and muddy picture with clear artifacts).

I hope that the FFmpeg developers can find a solution to this and that the solution then get ported over to MPlayer (and XBMC).

PS! Both XBMC and FFdshow uses FFmpeg libavcodec to decode XviD (however MPlayer uses native XviD)

[Heini011]

Hi,

many thanks Sharktooth for your great work!

i was satisfied with the EQM V3HR matrix, wich offers good quality and compressibility, for long time. but sadly it softed the images a little too much for my already denoised videos. so i created a modification with less hf and inter-frame compression but only a little less over-all compressibility.

give it a try...


08 10 10 10 11 11 13 15
10 10 10 10 11 12 14 16
10 10 11 11 13 14 16 18
10 10 11 13 15 17 19 22
11 11 13 15 17 20 23 27
11 12 14 17 20 24 30 36
13 14 16 19 23 30 40 48
15 16 18 22 27 36 48 62

13 14 14 14 15 16 18 21
14 14 14 14 16 17 20 22
14 14 15 16 18 20 22 25
14 14 16 18 21 23 26 30
15 16 18 21 24 27 31 36
16 17 20 23 27 32 40 48
18 20 22 26 31 40 53 64
21 22 25 30 36 48 64 83


greetings.

[Sharktooth]

Hi,

many thanks Sharktooth for your great work!

i was satisfied with the EQM V3HR matrix, wich offers good quality and compressibility, for long time. but sadly it softed the images a little too much for my already denoised videos. so i created a modification with less hf and inter-frame compression but only a little less over-all compressibility.

...

greetings.
v3HR is a HVS matrix. this modified version "lost" the HVS optimization and some compressibility. Yes, it retains more fine details but it's placed between v3HR and v3UHR. I was also thinking to "kill" both LR and HR and create a "midrange" general purpouse matrix no longer based on HVS studies by IEEE.
However feel free to mod any of my matrices, and let us know if you obtain better results.

[Heini011]

Hi Sharktooth,

>it's placed between v3HR and v3UHR

yes, but it gives only 9% increase in file size (@q=2) against v3HR, whereas v3UHR is already at 60%, which is too much for me.

i'am very satisfied with the results of my modded v3HR on denoised video material.

greetings.

[gumballguy]

Do the people in this thread have experience with HDTV encodes and these custom matrices? Ive done dvds (with standard quant matrices) and know what to expect there.... but its a little different when youve got bitrate ~6000, but a resolution in the vicinity of ~1800*1000 and are turning interlaced to progressive.
All numbers approx only, say rounded to 100's/1000's.

The high bitrate ordinarily would suggest V3UHR i would assume... but with such an increase in resolution?

[Sharktooth]

UHR? I suggest using LOW BITRATE matrices unless you're not scared by reaching bitrate as high as 6-8000kbps...

[Didée]

... and in case gumballguy is american and used to consult a "2nd opinion":

Most times, "low bitrate" matrices are the ones to use when encoding HD resolutions, indeed.

[gumballguy]

Lol. I was giong by GK, in which I'd have bitrates of 3000 to fit it to half a dvdr, or 6000 to fit it to a whole dvdr.

If you say low bitrate, well, I'll try it!

Would you suggest I start with EQM V3LR ?

[Sharktooth]

Try both LR and ULR. I'd also try 6of9 HVS. Jawor's 1CD matrix could be a good choiche too but IMHO it's too smooth.

[Backflip]

Should I assume properly mastered DVD's don't need to have additional bars on left and right of the picture? Bride & Prejudice has a rather thick bar on the left, and it messes everything up in terms of encoding :/

Requesting an opinion, but Sharktooth, what pixel/res/comp % do you aim for, in terms of the best utilisation of your matrices? Would you go down as far as a resolution of 576x448 (1.85:1). I currently aim for 80% and a 2.70 pixel (with 1 b-frame usually). Lower than 74% and I change matrix.

edit - should give some stat's. Currently encoding a familiar movie - Shark Tale (1.85:1).

576x448 (w/ 16:9 set in codec); next 0% AR error down from 720x560 (1.85:1)

Aim for 2 CD's @ average bit rate of 1890
All extra options off except for Cartoon Mode. Using Lanczos resize:
HR.V3 - 0.293, 54% of 0.538
LR.V3 - 0.293, 68.4 of 0.428
ULR.V3 - 0.293, 73.6 of 0.398

[Sharktooth]

well... black bars usually screw the final quality... crop them out without mercy :)
I usually try to keep the original resolution and choose the matrix according to matrix characteristics.
Compressibility test may tell you how much a matrix compresses but not what visual quality you will get from that matrix.
I know all my matrices (except EQM V2) are ultra sharp, so if i want a sharp encode i get one of them keeping an eye to the compressibility. If i want a soft encode i choose jawor's 1CD or EQM V2 (high-bitrate), or other soft matrices.

[Backflip]

Gonna try compress Shark Tale (1890 Kbps average)
HR.V3
2 b-frames and AQ
Comp test of 0.188 with 54.1% of 0.347.

[Backflip]

Having a look at the encode now, quite impressive :) One thing though - AR. Should it be indicated as 1.77:1 in MPlayer for a 1.85:1 movie? I used X: 16 and Y: 9 in the XviD codec settings, then 1:1 option in Gordian Knot. MPlayer prescales to 994x560.

After cropping just down to 720x560 I do have black borders, but it looks like if I cropped to 720x544 has absolutely no black border (W-Zoom @ 100% Y-Zoom @ 99% tho). Aspect error @ 1.4%. Does this matter or will the AR (X: 16 Y: 9) setting I use give perfect AR anyway? With a slight loss in quality because of stretching?

Thanks :)

[Sharktooth]

You should try to keep the source AR and crop away all the black borders.
Also, if you want a better upscaling (during playback) and you have a video card with Pixel Shader 2.0 capabilities, use Media Player Classic by gabest.
In the options go to Playback -> output. Set Directshow Video to "VMR9 (renderless)".
Now under VMR7/9 (renderless) pane set "Use texture surfaces and render video in 3D" then set Resizing to Bicubic A=xxx.

[Backflip]

Does keeping the source AR mean going into XviD settings and specifying AR of 1.85:1 (X: 185 Y: 100) with a movie which is stated to have an AR of 1.85:1 on the box. This was my original intention, but I just thought the setting X: 16 Y: 9 would do the same thing :(

[Sharktooth]

Yes, you should specify 1.85:1.
16:9 is 1.77:1 and that's a wrong aspect ratio for your movie

[Backflip]

Sweet, I think I got it right :) Cropped 12 from top and bottom, giving 720x544 (1.85:1/1004x544 on playback). No black left. I'm quite impressed with the encode. Only thing that really stands out as a minor thing is the streaking during the opening sequence (Dreamworks Blue background sequence).

I think I'm going to finally encode Bride & Prejudice (IMDB link - http://imdb.com/title/tt0361411/). Cropping away most black bars gives me either 720x432 @ W: 101% H: 100% (with black border on the left - clockwise: 72,0,72,0) or 704x432 @ W: 98% H: 100% (no black - 72,0,72,4). So I must go with 704x432 I believe, and give X: 235 Y: 100 in XviD codec settings (2.35:1 AR).

@ a bitrate of 1384 Kbps (2 CD backup) and being a musical with some high motion (dancing) I will probably use EQM V3LR which is returning (defaults & MSP:6, VHQ:4, Lanczos resize):

0.182 with 46.6% of 0.391
0.182 with 61.1% of 0.298 (1 b-frame) think I'll go with this one
0.182 with 63.2% of 0.288 (2 b-frame)

[JarrettH]

I haven't posted here forever, but I check these forums for xvid updates (isn't that sad). Thanks for your continuing work Sharktooth :)

[Sharktooth]

I will create another matrix (mid-range general purpouse) before "dropping" MPEG-4 ASP custom matrices.
I think there is nothing more to do coz there are plenty of good matrices out there.

[Backflip]

You should try to keep the source AR and crop away all the black borders.
Also, if you want a better upscaling (during playback) and you have a video card with Pixel Shader 2.0 capabilities, use Media Player Classic by gabest.
In the options go to Playback -> output. Set Directshow Video to "VMR9 (renderless)".
Now under VMR7/9 (renderless) pane set "Use texture surfaces and render video in 3D" then set Resizing to Bicubic A=xxx.

I'm not too sure I have Pixel Shader 2.0 capabilities, but I certainly noticed a smoother texture when I changed the settings. Using celtic_druid 's latest FFDShow for decoding. What should I use for A=???, .060, 0.75 or 1.00?

I also have noticed MPlayer (latest celtic_druid build) doesn't seem to upscale as well as MPC (with VMR9 (renderless) setting changed). Is there anything similar that you can switch in MPlayer so it upscales better?

Might post some pic's just to show how much smoother the picture is with the VMR9 (renderless) setting changed.

[Sharktooth]

uhm... dont know if mplayer supports VMR. however the sharpness in MPC is due to the bicubic resize (opposing to standar bilinear used by almost all players).

[Backflip]

Cool, now, I think I've got the whole anamorphic thing down. Just to go over it again - with 2.35:1 -- crop 72 from top and bottom, same with 2.40:1 material. Now, what is recommended for 1.85:1 material.

Also, what happens with video which is fullscreen, yet has bars (letterboxed?) added to the top? Can I still use the 1:1 options, then crop out the bars -- top and bottom (how much cropping should be needed?)? What Aspect ratio, the the XviD codec, would I give it?

Can't wait to try out the gerneral purpose CQM :)

[g8o4lyf]

Hello. I just finished making a 1/2 dvdr size xvid rip from spiderman 2. The original source was 1080i with ac3 audio and I ripped it to 2200 mb with the help of v3hr matrix. The main problem I faced was deciding on what resolution to choose for the rip. I checked out the sinbad xvid rip from cmore hd and i think the resolution for that was 832 by something. It still looked amazing in my friends hdtv. I wanted to avoid small resolutions because then after 14 hours or so of encoding its pretty pointless isnt it ? Ripping from 1080i to 6xx by xxx to me is same like making a 2cd rip from a dvd. So after compressibility tests and messing around I finally settled for the following:

Frame Size: 880 x 400 @ 29.970fps
Data Rate: 1998 kbps avg.(0.29% overhead) gknot showed 1897
Audio: Dolby ac3 384 kbps.

Here are some shots from my rip below. Shot 1 is to show the fire effects, shot 2 is the spidey and octavius fight on the building, if you look carefully you can see the web on docs face. Shot 3 is a fast motion scene when spidey is flinged from one building to another and the last shot 4 is my favorite, it shows water, fire, metal and explosion all at once. You might notice some artefacts in the pic because I had to save that shot at high quality instead of maximum as I cant upload more than 200 kb file here.

I am really pleased with my results and I cant say that I see artifacts in my rip because I dont. There are some parts with artifacts in the source itself so I have them in my rip but they are negligible.

If anyone has any ideas or pointers on how I could get a bigger resolution around this bitrate or any other optimaztions please share. This is the first time I ripped an hdtv ts file to xvid so I am sure there are lots of things to learn. Peace and thanks for providing the custom matrices.

[fight2win]

hi, i'm sorry if this question has been asked before, but pls can someone give me a link for a custom matrix for high quality 2-cd xvid rip of a dvd movie! thanks in advance!