PDA

View Full Version : i've made some tests


Q-Panic
1st December 2001, 01:34
i encoded edward scissorhands (pal) in some different ways... i always used the very low bitrate patch...

1. 4-passes, image quality priority 10: took about 11h... when the pictures are bright everything seems to be fine... but there are strange vertical lines at the edge of people and things... so i thought (it was a progressive pal dvd) i will deinterlace next time... even with 4-passes there were blocks on scence where edward cutted plants VERY fast (and the 20ths century fox logo in the beginning was blocky, too, because it's snowing heavily on it...) ;)

2. 3-passes, image quality priority 2: took about 12h because i used smartdeinterlace... which didn't improve the vertical lines at all, they were still there...:mad: picture quality pretty much the same... also the blocks...

3. cbr with image quality priority 2 and still WITHOUT temporal smoother (like above)... i used verticalreduceby2 (pal) to get rid of the vertical lines (just sounded good for it...:p )... but it wasn't...:rolleyes: ... well, i can live with those lines anyway... the quality was outstanding... even as good as the 4-pass... no, i am not kidding... although slightly more blocks in heavy motion scences... but it took only 4 hours so i don't really see the point of encoding 11 h when i get 90% of the quality in less then half of the time...:(

4. cbr like above just WITH temporal smoother this time... and you won't believe it but IT GOT WORSE! :eek: the temporal smoother made the picture kinda jerky and "swimming" (i mean it kinda moved... i can hardly explain it) and even worse: the credits got blurry on the edges which they hadn't been before... :mad: so i am really not a huge fan of ts right now... and i think everyone who had problems with blurry edges should try to encode without ts... in my opinion (and i think opinions here are devided about that) ts makes the picture WORSE...

no i will go to bed and while i am sleeping i will run a new test... this time i will use the the 1-pass vbr WITHOUT ts and i might use the smooth cg patch because i have read in this forum that it would do a great deal to the quality... i'll try... ;)

i will tell ya... if anyone wants to know it anyways... :p

later.
Q-Panic.

The Link
1st December 2001, 11:44
i will use the the 1-pass vbr WITHOUT ts and i might use the smooth cg patch because i have read in this forum that it would do a great deal to the quality... i'll try...
The "smooth cg patch"? What is it?

Q-Panic
1st December 2001, 14:27
you can download the cce patcher and with this little tool you can change the matrice which cce uses to encode videostreams... there are a few matrices which you can choose and play with... ;) this following extract is taken from a read me file which came with the patcher:

Each set has the following characteristics:

a) 'Standard' is the standard setting of Cinema Craft Encoder SP.
This setting was used for evaluation when the
MPEG-2 international standard algorithm was determined.
For natural images, this setting is appropriate.

b) 'MPEG standard' is the default setting of the
MPEG-2 international standard. If relatively high bitrate is used
and the image quality must be improved, use of this setting may
improve the image quality.

c) 'Smooth (CG etc.)' is appropriate for CG images (Computer Graphics) with low noise.

d) 'Very low bitrate' is intended to be used when the bitrate is
4 Mbps or less.

e) 'Ultra low bitrate' is intended to be used when the bitrate is
2 Mbps or less. But how low you can specify depends on source file.

f) 'Angel SVCD 1CD 2.35 avg800 max1500'


at some other place in this forum somone recommended the smooth cg... so i gave it a try... (;

cheers.
Q-Panic

The Link
1st December 2001, 15:16
Thanks!!! :)

Q-Panic
1st December 2001, 21:05
ok... i now encoded 15minutes which is about 115minutes long:

1. 1-pass vbr (avg. bitrate 1600) and smooth cg matrice, iqp 2
well, the picture quality in was not so good, but ok when it didn't move... but even during normal motion it showed blocks... at least too much for me.... :rolleyes:

2. cbr (avg. bitrate 1600) and very low bitrate, iqp 2
the picture quality was good, but during motion it became blocky... but all in all MUCH better than the 1-pass vbr... although i do believe that the only way to get rid of the blocks will be a 4 pass... i got no blocks in chocolat which is almost as long as 15 minutes and i encoded it with a 4 pass... ;)

BUT the picture was jerky during, especially during camera movements... and it WASN'T jerky after the 1 try... conclusion: the very low bitrate makes the image JERKY!!! :mad: so don't use it... it is worse for PAL movies, though, because i encoded some NTSC movies aswell with this matrice and this side effect didn't appear so obviously...

so what's the moral of the story:

1. for movies up to 106 minutes (the avg. bitrate will stay above 1800!!) it is OK to use the CBR... there are slightly more blocks than with a 4-pass but it takes only a 1/4 of time... :p

2. for movies which are longer then 105 (and shorter than 118) use mulitpass with at least 3 passes... because the avg bitrate will then fall under 1800 and this will result in more and more blocks, depending how long the movie actually is...

3. for movies over 118 minutes you better use 3 cds... :D

4. don't ever use cce's matrix patch "very low bitrate" for encoding PAL videos... :mad:

5. the "smooth cg" matrix patch seems to be very good... :)

6. i have to go now because it's 9 a.m. and i should have been on a party since 8... :(

bye.
Q-Panic.

CrazyPortuguese
25th January 2002, 13:17
Party's at 8 AM :)
Great job analizing the matrixes it seems that smooth cg patch is the way to go, right now I'm gonna try the Tmpegenc cg patch to see how it works with 2 hour movies on two 80 minuts cd's.