Reloaded HQ encode !!!

[Teegedeck]

OK, I re-did the snapshots. BTW, encoding Matrix Reloaded with SixOfNine at quant=2 (anamorphically) would yield about 3.2 GB, semi-insane at quant=2 about 4.7 GB.

[Soulhunter]

Originally posted by Teegedeck
OK, I re-did the snapshots. BTW, encoding Matrix Reloaded with SixOfNine at quant=2 (anamorphically) would yield about 3.2 GB, semi-insane at quant=2 about 4.7 GB. Nice... :)

Btw, my CM @ quant 2 gave 6.5GB (with 384kbps AC3 stream) !!!


Bye

[BoNz1]

Originally posted by CruNcher
take this
http://cruncher.mufflastig.com/XviD/samples/
:devil:

Nice, and not just the video 8D. I did Matrix Reloaded onto 2 cds with AAC audio IIRC at 704x304 res or something like that. I was reasonably satisfied with the quality.

[atreya2011]

Originally posted by Teegedeck
Comes with LigH's matrix editor.

Sorry, but I downloaded LigH's matrix editor and I could not find the semi insane matrix. Is there any other place I can get it from?

[JasonFly]

Soulhunter, did you try IIP scrip^t of avisynth. I don't know many things about that script but it seems that his goal is to upsize the picture to 1024*xxx.

[Jawor]

@atreya2011: try here (http://www.ligh.de/software/qmatrix.zip)

[Didée]

Originally posted by JasonFly
it seems that his goal is to upsize the picture to 1024*xxx. No - the (initial) idea was to upsize to 1280*720 (720p). But it doesn't matter, it works with any output resolution.

And yes - Soulhunter already tried that script. It gave him already alot of reasons for ":rolleyes:"

;)

- Didée

[P0l1m0rph1c]

Originally posted by atreya2011
Sorry, but I downloaded LigH's matrix editor and I could not find the semi insane matrix. Is there any other place I can get it from?

http://dc0der.is-a-geek.org/qmatrix.rar

[Soulhunter]

Originally posted by Didée
It gave him already alot of reasons for ":rolleyes:"
Yeah, results where superb... :)

If you dont care much about speed its a wonderful thing !!!

Maybe I do a full IIP encode after the SharpResize one too... :rolleyes:


Bye

[SeeMoreDigital]

This is great stuff. I can't believe this thread passed me by!

Soulhunter I'm glad you've carried on experimenting with 1024x576 'true 16:9 frame' PAL encoding, since we last talked about it all those months ago.

I would be interested to learn if any NTSC coders have had any success generating encodes with 864x480 'true 16:9 frame'? All the ones I generated (from sections of T2 Extreme and StarWars 1) looked excellent and worked very well on my TV via hardware playback card but I did not have much luck at all playing them back via software player.

Anyway you've certainly generated a lot of interest in the subject.

Cheers

EDIT: I forgot to mention. If there are NTSC users interested in generating 'cropped and resized' encodes from 2.35:1 or 2.40:1 DVD sources, you might find using 864x368 and 864x352 image pixel frame sizes (respectively) provides a closer match to the DVD.

[CruNcher]

you could upsize and then use this here but be carefull it's slow

http://www.compression.ru/video/smart_sharpen/index_en.html

[Soulhunter]

Originally posted by SeeMoreDigital
Soulhunter I'm glad you've carried on experimenting with 1024x576 'true 16:9 frame' PAL encoding, since we last talked about it all those months ago. To tell truth I was a bit skeptical first... :rolleyes:

But now I have to gratitude you for forcing me to use such framsizes !!!


Originally posted by CruNcher
You could upsize and then use this here but be careful it's slow Thanks !!!

I will have a look at this... :)


Bye

[JasonFly]

It a little off topic but i'm wondering if such resolutions(1024*xxx) are compatible with standalone players.
I'm not sure I have make me clearly understood, so I give you some more info:
I don't know if Soulhunter XviD options are standalone compatible, but imagine he has chosen 1 bframe and no other "exotic" option, does this resolution can be play on a standalone? That is the question.

[Gaia]

Originally posted by JasonFly
It a little off topic but i'm wondering if such resolutions(1024*xxx) are compatible with standalone players.
I'm not sure I have make me clearly understood, so I give you some more info:
I don't know if Soulhunter XviD options are standalone compatible, but imagine he has chosen 1 bframe and no other "exotic" option, does this resolution can be play on a standalone? That is the question.

Most of the standalone players have big problems decoding high bitrate dvd resolution encodes. 1024x is definetly far too much for any player.

[Soulhunter]

Yeah, second that !!!


Even when there wouldn't be problems with the relative high resolution...

...the bitrate would be much too high for the current standalone chipset's !!!


Quite not sure if high resolution but "spec compliant" bitrate would be playable... :confused:


Btw, Ive only used 1 B-VOP !!!

See my settings in the first post... ;)


Bye

[SeeMoreDigital]

Originally posted by Gaia
Most of the standalone players have big problems decoding high bitrate dvd resolution encodes. 1024x is definitely far too much for any player. Yes, unfortunately current hardware player chipsets can only cope with a maximum image pixel frame size of 720x576 (total 414,720 pixels). Worse still most hardware players stick to set pixel sizes/ratios meaning experimentation is limited!

I would have hoped that for you NTSC hardware player guys, would have been able to use say frame sizes of 864x480 (414,720 total pixels) but sadly this does not seem to be the case (Sigma's Xcard can though :D )

The main advantage of using such wide ratios becomes quite clear when viewing very quick panning shots and any high motion scenes. This has got to be due to having much more horizontal pixels (20% more for NTSC and 42% more for PAL) in the frame!

Anyway, for all those who have not tried yet... please give it a go!

Cheers

EDIT: Good news for all you NTSC users. If you have DirectX 9 installed and adjust your software media players playback colour out put to VMR9 'windowed', it is possible to view 864x480 encodes (or their cropped equivalent) without any video noise appearing at the far right of the frame.

[Didée]

Originally posted by CruNcher
you could upsize and then use this here but be carefull it's slow

http://www.compression.ru/video/smart_sharpen/index_en.html
Looks interesting ... but they seem to offer only those preview pics. Is there a download "hidden" somewhere in the russian pages?

However, from eyeballing these samples, it looks like this sharpener uses some sort of adaptive/gradualual unsharp-masking. Like "for each pixel, search the max. distance over which similar pixels occur continuously, then use that as the radius for unsharp masking this one pixel". Something like that.

I took the opportunity of processing one of their examples with iiP ;)

Have a look here:Comparison_MSUSmartSharp--iiP.zip (http://kotisivu.mtv3.fi/ii00i/Comparison_MSUSmartSharp--iiP.zip) (ZIP 138kB)

(Note: the artefacts in the iiP-sample are already in the source (jpeg :rolleyes: ) - the source was feeded as-is into iiP, without any pre-processing).

The MSU filter does a great job in contrast enhancement, clearly better than iiP in this respect.
But taken literally, contrast enhancement is not "sharpening", it's ... contrast enhancement. And thus, the result of this MSU filter looks a little "thick" or "greasy" -- the usual effect of unsharp-masking.

Decide for yourself which one gives the sharper image.


- Didée


P.S.

For the curious: settings for iiP wereiip(
\ duststr=2, dustweight=1.0,
\ ss1_x=1.414, ss1_y=1.414,
\ detailcontr1=127, detailcontr2=192, contr_radius=5, PixSharp=1.0,
\ ss2_x=3.5, ss2_y=3.5,
\ Xstren=255, Xlimit=255,
\ subpelstren=1.58, antiflicker=false,
\ flatweight=0,
\ protect_floor=0, protect_bias=12,
\ dering=100, dering_weight=1.0,
\ dering_floor=4, dering_bias=6,
\ debug="-showall", cropx=0, cropy=16
\ )

[Soulhunter]

Originally posted by Didée
Looks interesting ... but they seem to offer only those preview pics. Is there a download "hidden" somewhere in the russian pages?
Yeah, cant find the download link... :(


Bye

[CruNcher]

http://www.compression.ru/video/smart_sharpen/src/smart_sharpen.zip

[TheUnforgiven]

good link CrunCher :D
is there a chance that u post a link to the deinterlacer if there is.
sorry for the offtopic

[CruNcher]

it's CruNcher ;) no sorry the Samsung ones are not public as it seemes

[TheUnforgiven]

used to write it CrunCher as its the name of my pc :cool:
thanx anyway.

[Soulhunter]

Originally posted by CruNcher
http://www.compression.ru/video/smart_sharpen/src/smart_sharpen.zip Thanks !!!

Btw, can you speak/read russian lingo ???


Bye

[Soulhunter]

Just to fill the time till my SharpResize encode is done...


Here are my custom covers for the JewelCase !!!

Front:
http://img8.imageshack.us/img8/6028/Matrix2_cover02c.jpg

Inlay:
http://img8.imageshack.us/img8/7954/Reloaded_Inlay.jpg

Back:
http://img8.imageshack.us/img8/5580/Reloaded_Back.jpg


Bye

[nanogate]

When i try to get over 640x480 with sharp resize Virtualdub shows Avisynth readerror.
But its working with Lanczosresize.:confused:
Do I have an old version of Sharptools? (v0.1)
Its a little slow on my machine@640,352 firstpass (1-2 fps)
With Lanczos i got ~15 fps.
No other Filtering.

[SoonUDie]

@nanogate
SharpReszie is definately not meant to be fast!

At the moment, almost all DVD->HDTV conversion efforts are more "experimental" than practical, considering the speed loss... unless of course you enjoy encodes that take a couple days to finish!

Also, for your read error, Avisynth probably ran out of memory. To fix this, put the line

SetMemoryMax(amount)

at the beginning of your .avs file, where "amount" is the amount of memory you want to devote to the task (in general I use 128+).

For anyone else thinking of DVD->HDTV conversions, you should try fastEDIUpsizer (but this is also veeeery slow!), posted a while ago in this forum. It uses edge detection to get a more more pleasing (jaggie-free!) upsize than Lanczos or Bicubic.

[nanogate]

lol I often use setmemorymax(96) but dont use it when I need it :eek:

[SeeMoreDigital]

Originally posted by SoonUDie
...For anyone else thinking of DVD->HDTV conversions, you should try fastEDIUpsizer (but this is also veeeery slow!), posted a while ago in this forum. It uses edge detection to get a more more pleasing (jaggie-free!) upsize than Lanczos or Bicubic. What exactly do you mean when you say DVD->HDTV conversions?

If you are suggesting encoding using more vertical pixels than the DVD source (ie: more than 480 or 576) then in my opinion it's really not worth the effort!

I've done plenty of tests in this regard but came to the conclusion that only increasing the horizontal quantity of pixels is useful exercise!

Cheers

[Soulhunter]

Originally posted by nanogate
When i try to get over 640x480 with sharp resize Virtualdub shows Avisynth readerror.
But its working with Lanczosresize.:confused:
Its coz SharpResize uses super-sampling (4x IIRC) !!!

Means it resizes to 2560x1920, adds sharpening and resizes back to 640x480... ;)


Bye

[atreya2011]

Originally posted by SoonUDie
Also, for your read error, Avisynth probably ran out of memory. To fix this, put the line

SetMemoryMax(amount)

at the beginning of your .avs file, where "amount" is the amount of memory you want to devote to the task (in general I use 128+).



Sorry for being offtopic, but virtual dub has crashed many a time for me during encoding at random points and I know it is defenitely a memory problem, so I can Use SetMemoryMax to fix it right?(btw I have PC 3200 512 MB DDR)

[Soulhunter]

Originally posted by atreya2011
...so I can Use SetMemoryMax to fix it right?Just try it... ;)

Use also this tool (http://hcidesign.com/memtest/download.html) to test your RAM's stability !!!


Bye

[SoonUDie]

@atreya2011
Like Soulhunter said, try it and find out. I would assume that there is a good chance it will work for you.

@SeeMoreDigital
It's true that you can't really recreate information by upsizing using any known spatial technique. However, you can create *pleasant* results. Compare the SharpResize Matrix shots with the lanczos ones to see what I mean... the end result is that you get something much sharper that is more enjoyable to watch if you play it back on a high-definition display. Also, if you use something like fastEDIUpsizer, you can eliminate aliasing artifacts (jaggies).

[Teegedeck]

The more sane option would be to do this via postprocessing though... Either sharpening in ffdshow could have a similar effect (haven't tried it) or a sharp resizer used during playback.

[SoonUDie]

@Teegedeck
I think the point was that sharp resizing isn't something you *can* do in realtime with our current software... so you have to encode it that way. Otherwise, the only other option I can think of is a hardware upconversion box.

[Teegedeck]

Truely, 'sane' was the wrong expression when we reach hardware limitations here; 'space-effective' says it better. It is a niche thing, but upscaling while encoding is justified when the outcome looks nice and storage is cheap, I think.

[SeeMoreDigital]

Originally posted by SoonUDie
...SeeMoreDigital
It's true that you can't really recreate information by upsizing using any known spatial technique. However, you can create *pleasant* results. Compare the SharpResize Matrix shots with the lanczos ones to see what I mean... the end result is that you get something much sharper that is more enjoyable to watch if you play it back on a high-definition display. Also, if you use something like fastEDIUpsizer, you can eliminate aliasing artifacts (jaggies). When generating encodes from an 2.35:1 or 2.40:1 source, what vertical image pixel frame sizes have you tried?

Cheers

[Soulhunter]

Yay, its done !!!

I just finished my SharpResize version... :D


Source: Matrix Reloaded (PAL/R2)

Image: 1024x576 Pix. @ 25fps | 132 min.

Settings: XviD v1 RC4 | 2Pass (VHQ1) @ 4199963 KB | BVOP's = 1 / 1 / 1 | Matrix = Soulhunters v3

Audio: Dolby Digital (AC3) | 5.1 Channels | 384 kBit/s

Time: About 4800 min. (time to encode) ("2600XP")


AviSynth script:

Clip = Last

Clip1 = Clip.VagueDenoiser(Threshold = 0.25, Method = 1, NSteps = 6, Chroma = False)
Clip2 = Clip.VagueDenoiser(Threshold = 1.00, Method = 1, NSteps = 6, Chroma = True )

Clip1.MergeChroma(Clip2)

SharpResize(1024, 576)


Samples:

http://img9.imageshack.us/img9/5696/Reloade_SR_127666.jpg

http://img10.imageshack.us/img10/3161/Reloade_SR_129235.jpg

http://img10.imageshack.us/img10/2896/Reloade_SR_139423.jpg

http://img9.imageshack.us/img9/8545/Reloade_SR_169657.jpg

http://img9.imageshack.us/img9/3371/Reloade_SR_170930.jpg


Bye

[JasonFly]

Your screenshots are amazing! Especially the last one with the explosion.I haven't compared with the previous but these are impressive.
Congratulations for your patience doing all these tests.(I can't imagine encoding time if you had set vhq to 4. But maybe the low speed should be
due to sharp resize? I don't know this filter very well)

What I must remember for these tests is that upsizing is definitely the solution to reduce(I should say: not to introduce) artifacts.This of course implies DVD storage but that's worth the quality I think

[Soulhunter]

Originally posted by JasonFly
But maybe the low speed should be due to sharp resize?Ehm, mostly... :rolleyes:


Source:

The Matrix - Lobby Shootout / 3:07 min. @ 25fps / 1024x576 pix.


Settings:

XviD v1 RC4 (Koepis)

Fixed Quant 2 (VHQ1)

I-frame interval = 250

Matrix = Soulhunter v3

BVOP's = 1 / 1 / 1


Results:

LanczosResize
Size = 154 MB / Time = 05 min.

SharpResize
Size = 162 MB / Time = 65 min.


Bye

[bot]

I know you're either going to hate me or ignore me for asking this.

-But since transcoding the lossy dvd to divx is bound to introduce artefacts and lessen pq in other ways... and because of the 80 hour encoding... what kind of result would you get from resizing the source dvd during playback (with the highest possible quality resizer available at real time)?

I can't do the comparison myself for two reasons. One is that I don't have Matrix Revolution DVD (which is the frame of reference here). And the other is that I can't make the simplest avisynth script without getting OutOfMemory.

[kilg0r3]

Originally posted by bot
what kind of result would you get from resizing the source dvd during playback (with the highest possible quality resizer available at real time)?The disadvantages are (a) the macor blocks larger compared to the overall resolution and (b), which is related to (a), that you resize (enlarge) the artifacts introduced by recompression. And the other is that I can't make the simplest avisynth script without getting OutOfMemory. [/B]
SetMemoryMax(), IIRC, might help you there. But check the exact syntax

[bot]

Originally posted by kilg0r3
The disadvantages are (a) the macor blocks larger compared to the overall resolution and (b), which is related to (a), that you resize (enlarge) the artifacts introduced by recompression.
SetMemoryMax(), IIRC, might help you there. But check the exact syntax


A) The macro blocks of a dvd aren't that visible compared to in xvid (or any other) transcoding?

B) Recompression? I know DVD's are compressed, but that's a little bit harshly said?


Tried SetMemoryMax(). When I uninstalled Avisynth and reinstalled it (latest version) I stopped getting Outofmemory. Now the player (mpc) just switches off. It's noob-o-rama here :rolleyes:

[Teegedeck]

Wow, great screenshots! Congrats for that nice encoding, Soulhunter. :)

@kilg0r3: I doubt that macroblocks are of importance in this context; an anamorphic encode takes less bits than a true 16:9 frame and thus is less likely to introduce new 'blocking'. (For encoding it should be avoided that macroblock borders become visible or artifacts appear IMHO.)

[Soulhunter]

Originally posted by bot
I know you're either going to hate me or ignore me for asking this. No fear, I will only make my usually sarcastic comments... :D


Originally posted by bot
But since transcoding the lossy dvd to DivX is bound to introduce artefact's and lessen pq in other ways...Most of the visible artefact's are already in the DVD source !!!

And btw, this here is the XviD forum, not the DivX forum... ;)


Originally posted by bot
What kind of result would you get from resizing the source dvd during playback...Ok, the source realtime-resized with Lanczos would look nearly identical to my 1st encode... :rolleyes:

But the source files would need more than 5GB storage IIRC !!!

So there is no chance to backup them on a single DVD5... ;)


Originally posted by bot
I can't do the comparison myself for two reasons. One is that I don't have Matrix Revolution DVD (which is the frame of reference here).Nope, its Matrix Reloaded !!!


Bye

[bot]

Originally posted by Teegedeck
@kilg0r3: I doubt that macroblocks are of importance in this context; an anamorphic encode takes less bits than a true 16:9 frame and thus is less likely to introduce new 'blocking'. (For encoding it should be avoided that macroblock borders become visible or artifacts appear IMHO.) [/B]

Using PAL for example:
Are you saying that backing up in anamorph xvid 720x576 creates more artefacts than resizing to 720x384?


@soulhunter ... *double sided migraine* Of course I meant xvid and reloadad. :scared:

But the source files would need more than 5GB storage IIRC

-I meant making an avs file, referring to the untouched dvd, resizing it in real time:
DVD minus extras + 10kB avs < Original Size

Edit: I get the sinking feeling that I'm missing a point.

[Teegedeck]

Originally posted by bot
Are you saying that backing up in anamorph xvid 720x576 creates more artefacts than resizing to 720x384?

That's a bit of a misunderstanding indeed. Soulhunter has resized his encoding to 1024x576; so this takes indeed more bits than an anamorphic 720x576. I even cropped off the black barrs and encoded 704x416 anamorphically. And that naturally saved me quite some bits when compared to a 'true 16:9 frame'.

Yes, on the other hand you could say, if you go lower and lower with the bitrate, that a 720x576 encode would probably start looking bad sooner than a 720x384 encode. Or not; anamorphic encodes tend to mask that effect pretty well. I'd say it depends. If you go to the extremely low bitrates, anamorphic encodes will look worse than non-anamorphic ones unless you encode the anamorphic frame at an altogether lower resolution than the non-anamorphic frame. But that's not the sort of thing we were talking about.

Neither of us got any artefacts in his encoding. We don't do such things. :cool:

[SeeMoreDigital]

Originally posted by Teegedeck
...I even cropped off the black barrs and encoded 704x416 anamorphically. And that naturally saved me quite some bits when compared to a 'true 16:9 frame'. Have you tried encoding a 2.35:1 DVD source to an resized 1024x432 frame yet?

Cheers

[bot]

(For encoding it should be avoided that macroblock borders become visible or artifacts appear IMHO.)

So here you are referring to... what? Keeping 720x576 is bad visavi Resizing to 1024x576?








/resisting urge to hijack the thread by following up on the 720x324 :)

[Teegedeck]

I'm not talking about anything that specific, just about using a high-enough bitrate! That's all.

[Teegedeck]

Originally posted by SeeMoreDigital
Have you tried encoding a 2.35:1 DVD source to an resized 1024x432 frame yet?Cheers
Err, no. I'm sorry but ATM I don't see what you wanted to say with that.