XviD and "noisy" movies

[RebelYell]

Hi there,

I just wanted to let you know guys - my first date with XviD ;-)
Anyone ever tried to rip some old DVD "noisy" movie? Well, after some
jerky results with 3.11 SBC and 4.12 and even 5 Pro Bundle i decided to look for XviD.
The movie i'm talkin' about is Blade Runner - The Director's Cut.
Yes, it's pretty damn noisy, with tipical "old cinema" artifacts and
no matter what bitrate/resolution/ nr of CDs i've tried - the result
was jerky...
I mean, the movie is pretty dark, and all that "cinema" noise was
turning into blocks... even at 1900kbps...
I decided to look for something else, and just found XviD.
Well, a 2 CD rip looks great... :)
And what is so cool, that "noise" is still there, like on the DVD
not being translated in blocks, or blurry / faded scenes.
It looks almost the same as the DVD...

This is what i call a kewl piece of software...

If anyone's interested i'm using Gknot. (and of course the DeeNine's guides ;)

Over and Out,
Rebel Yell

[kagoru]

I did Scarface, a movie which has a lot of noise! I encoded it using divx3.1, divx 4, divx5(all the xtras turned on) and xvid!

I think the divx5 rip looks pretty good! pretty good picture and no artifacts. Xvid is a little bit inferior compared to the divx5 rip and has to fight with the noise.

[Gazza]

Maybe its a case of Xvid suits some movies whereas Divx5 suits others?

I tried encoding 'Finding Forrester' with both Xvid and Divx5 to judge for myself what looked the best. I used Doom9's excellent guidlines as the basis for my experiment. I noticed that even though Divx5 seemed to have slightly brighter colours (speculative guess) in the color intensive scenes I still go for Xvid. Why? In the Divx5 encoded output, I noticed that were there is a dark object against a bright background there was shifting horizontal blocks of pixels when the dark object moved. I thought at first it could be my graphics card (which must play a huge part in the quality of any encoded movie). But I then thought not, as I didn't see the same effect with the Xvid encoded output.

Just thought I would share this with everyone. I would be interested to know if anyone else wants to run the same experiment and confirms the same effect with the same movie?

Gazza

[Gazza]

Before anyone asks I had Qpel turned off.

Gazza

[canadian_fbi]

just to add my two cents... i recently did rips of the godfather parts 1 and 2. 2 cd, ogm rips at 175 and 200 minutes, respectively. being old movies, there was lots of grain and noise in the picture, and the second movie actually came on two dvds since it was so long. because the movies were so long, the bits/pixel*frame ended up being about .200 on the godfather part 2. with all the noise and crap in the source material, i must say that the codec did a fantastic job giving a good picture at this low b/p*f ratio. noisy movies, clean movies, all movies bow before the mighty compression power of xvid :devil:

[kagoru]

What Resolution did you all use? I went for 640x272 on a 2h 43m movie; that gave me 0.230 bits/pixel*frame. Did you turn on any special switches with xvid?

EDIT: What Post Processing options do you use to compare?

[canadian_fbi]

i used 544x304. i think the end bits/pixel*frame was more like .210, actually. but if it weren't for the noisy source this would have been almost excessive. i was going through some of the old guides and reading the parts about .20 as a guide for 1 cd rips and .27 for 2 cd rips, and i was thinking how .200 is almost an upper bound for me lately. i think i went with .180 for a movie recently. it depends on how clean the source is and how much action there is of course, but these are pretty impressive numbers all the same, and it's xvid doing the magic i believe.

[Gazza]

When I did 'Finding Forrester' on 2 cd's Gknot gave me 672 x 288 with a bits/pixel*frame value of 0.274.

By the way, does anyone notice that the final file size for encoded animated films is considerably less than what the non-Gknot bitrate calculators estimate. That is, I have always been pretty close with 'real' films but animations have always been much smaller. It doesn't seem to happen so much when I recently started using Gknot.

[canadian_fbi]

the file size has been off with what codec, xvid? it's never given me any problems so far sticking to its desired size. in fact, it's come amazingly close sometimes (like within a few KB). makes me wonder how divx 4 could have ever been so bad undersizing files :)

[Gazza]

Sorry. This was with Xvid. But my earlier bitrate calculations where done using 'divx4bitrate.exe' or 'BitRateCalc.exe'. Using Gknot seems to give me the same order of success as you have experienced. Maybe the combination of Gknot and Xvid is the solution?

Gazza

[Teegedeck]

Hi,

I'm curious to learn if you have used the MPEG-quantizer or the h.263-quantizer on those movies (sounded like MPEG-quant to me).

Greetings,

Tee

[RebelYell]

h.263-quantizer for my Blade Runner...


Over and Out,
Rebel Yell

[RebelYell]

Anyone with MPEG?


Over and Out,
Rebel Yell

[sierrafoxtrot]

i remember encoding in the mood for love with H263, MPEG and mod Quant, and in all cases, the graininess of the original source was preserved with all three quantization methods ... maybe it's just the way the codec processess stuff.

[kagoru]

I went with MPEG following Doom9! But I guess I'll redo the rip using a lower resolution and h.263-quantizer because I'm not really satisfied with the result!

[Gazza]

I used a 2 CD rip with the following settings for the Xvid codec;
Motion search precision : 5
Quantization : MPEG
FourCC used : XVID
Lumi masking was not selected
Min I & P frame quantizer : 2
Max I & P frame quantizer : 31

The rest of the parameters where left at default as per Doom's guide. Would going to H.263 and maybe lumi masking be better choices for a 2 cd rip? I will give it a go for myself but I'm interested in hearing people's feedback.

Gazza

[OUTPinged_]

why not just filter the damn noise out?

(and dont look at me like that)


MooPolice approved rips use noise filtering, too.

[canadian_fbi]

Originally posted by OUTPinged_
MooPolice approved rips use noise filtering, too.
oh, the police are always telling me to keep the noise down ;)

[Teegedeck]

@Gazza:

Personally, I'd only use lumi-masking if you reach a very bad compressibility in GKnot (about 35%). And not in movies that have low contrasts/are dimm.

[Gazza]

@Teegedeck

With regard to lumi masking what you say makes sense once you understand what it could provide for you on a 2 cd rip. What are your thoughts on H.263 versus mpeg? I'm not completely up on what the low down technical differences are between these two so better to ask. I've done 2 cd rips with mpeg but will try soon with the H.263 to see if there is any noticeable difference.

The problem with comparing one against the other, is that if the difference is subtle then the human eye couldn't help anyway as humans are very adaptable creatures (reason for our success?).

Gazza

[serbersan]

I've made too a rip from an old movie, Fist Of Fury 1:42 min, It's a film of Bruce Lee.

The dvd was very noisy, with rain scenes and high action. The compressibility with Gknot divx3 was 19% at 640x256.

The 2 CD rip it's great, some little blurred areas but retain grainnes and noise incrediblely well.

Used modulated type with 2-4 quantizers.

One question, as I believe -h said, mpeg quantizer could introduce noise, and yes I've made two rips Fast & the furious, and Remember the titans, and in both cases the result was identical to dvd but with a little noise if you see near the screen.
Could the custom-matrix solve this problem?

Thanks

[RebelYell]

[QUOTE]Originally posted by OUTPinged_
why not just filter the damn noise out?

Hehe, well, i still wait for the Lucas Arts Tutz about how they re-mastered the old Star Wars cinema film, and after that i'll go buy their hardware and re-master my old BladeRunner DVD.
It's a DVD and a very old movie not a cinema screener!
Grain and film articfacts stuff...

:D

Over and Out,
Rebel Yell

[OUTPinged_]

ok, you may not filter it all, just apply some tempsoften with no-so-stupid threshs and it will help pyur movie to be more compressable.


i did the same thing for casino. stupid thing wanted me to encode it at 50% 1/2pass@ 640x256. and that was 3cd rip :-E tempsoften was transparent (not noticeable) and gave me 8% reduction in bitrate.

these r5 dvds suck.

[RebelYell]

Gee, i have the PAL version of Casino DVD...real pain in the *ss to encode. My output was always blurry, sort of un-focused...

I'll give it a try with XviD these days...


Over and Out,
Rebel Yell

[OUTPinged_]

heh, about that casino dvd again. switched it to --alt-preset standard mp3 with latest dll (came out at 190k, ALOT) and 1/2pass ratio jumped to 69%, which produced a ok looking rip.

i laugh at people who encoded this 3h movie as 512x 1cd rip and claim they have achieved "good" quality.

[kagoru]

Is the compressability the same as with divx3.1? Means: Do I have to stay above 65% do get decent quality?

[OUTPinged_]

well in general yes, but instead of getting blocky scenes you will have them with typical mpeg artifacts if you go lower.

it is ok to go under 50% for xvid, it will just get more blurry.

still, you want to have sharp picture, you go for 66% 1/2 pass ratio.

(66 imo is the best value in terms of "picture quality/compression" for divx311. it would be safe to go with xvid for high quantizers on hi-mo scenes (more compression) but it doesnt have motion detection :-)

on quantizers 2,3,4 it looks about the same as divx311. and compresses about the same (excluding that h23-something cheating, grr).