Avisynth: Latest Script Discussion - Page 2

Wolfi · #21 04-02-2003, 03:24 PM

I'm not using MovieStacker so can I leave that line out of my script (source_anamorphic=false)

//Wolfi

Kane · #22 04-02-2003, 03:27 PM

Quote:

Originally Posted by Wolfi

I'm not using MovieStacker so can I leave that line out of my script (source_anamorphic=false)

//Wolfi

perhaps you should *g*
makes things a lot easier, especially adjusting the aspect ratio to preven t "egg-heads"

if your source isn´t anamorphic, you can leave it out

Wolfi · #23 04-02-2003, 03:35 PM

I forgot... I dont know so much about all this filters so were can I find readme's for

GripCrop, GripSize, all the Dust filters for example Goldust, TemporalSmoother, TemporalCleaner, STMedianFilter

It's not easy to know witch comand I should use in different scripts, thats why I need to know were I can find readme's

I have found some readme's on this site but I'd like to know if there is readme's for the filters I counted up and not the once on this site

http://www.jungleweb.net/~sansgrip/avisynth/

//Wolfi

kwag · #24 04-02-2003, 03:38 PM

Quote:

Originally Posted by Wolfi

but I'd like to know if there is readme's for the filters I counted up and not the once on this site

You can find most of the links here: http://www.kvcd.net/forum/viewtopic.php?t=2553

-kwag

Kane · #25 04-02-2003, 03:43 PM

Quote:

Originally Posted by kwag

My file size difference using the temporal filters is ~150KB lower per every 30 seconds. Let me know your results

-kwag

did you use the same cq value for this comparison, or did you compare the final sample, ToK gave you?

Wolfi · #26 04-02-2003, 03:49 PM

Were can I download TemporalSmoother

//Wolfi

Kane · #27 04-02-2003, 03:55 PM

Quote:

Originally Posted by Wolfi

Were can I download TemporalSmoother

//Wolfi

i would suggest you download moviestacker
even, if you don´t use it, you´ve got the most common filters in the install-directory.
temporalsmoother is the same dll as temporalcleaner ( TemporalCelanerOld.dll )

kwag · #28 04-02-2003, 03:59 PM

Quote:

Originally Posted by Kane

Quote:

Originally Posted by kwag

My file size difference using the temporal filters is ~150KB lower per every 30 seconds. Let me know your results

-kwag

did you use the same cq value for this comparison, or did you compare the final sample, ToK gave you?

I didn't use ToK. I used a fixed CQ of 70 for both samples.

-kwag

Kane · #29 04-02-2003, 04:02 PM

well, then i´ve already posted it.
but i will re-try it in a few minutes, just to make sure

Kane · #30 04-02-2003, 04:47 PM

let´s see:

the script without temporalcleaner and smoother:

Code:

LegalClip()
GripCrop(528, 576, overscan=1, source_anamorphic=false)
GripSize(resizer="BicubicResize") 
SpaceDust() 
sharpen(1.0) 
mergechroma(blur(1.58)) 
mergeluma(blur(1.0)) 
GripBorders() 
LetterBox(16, 16, 16, 16) 
LegalClip()
Sampler(length=24)

size: 11,2MB
encoding time: 5:26
sample length 86sec
cq 70

the script with temporalcelaner and smoother

Code:

LegalClip()
GripCrop(528, 576, overscan=1, source_anamorphic=false)
GripSize(resizer="BicubicResize") 
SpaceDust() 
sharpen(1.0) 
TemporalSmoother(radius=2, strength=2) 
TemporalCleaner(ythresh=1, cthresh=15) 
mergechroma(blur(1.58)) 
mergeluma(blur(.7))
GripBorders() 
LetterBox(16, 16, 16, 16) 
LegalClip()
Sampler(length=24)

size: 11,3MB
encoding time: 7:34
sample length: 86sec
cq 70

maybe i´m too stupid to understand, but what should be the advantage of the script WITH temporalsmoother and cleaner? it takes longer to encode, but the image quality is the same, as without temporalcleaner and smoother.

kwag · #31 04-02-2003, 04:54 PM

Here's your problem Kane: mergeluma(blur(.7)) on one script and mergeluma(blur(1.0)) on the other. The tests I did were with mergeluma(blur(1.0)) on both. This way, you get smaller file size with the temporal filters, and the image looks the same. However, with the temporal filters and lowering the value of mergeluma to .7, we get a slightly sharper picture and about the same file size. Sorry if I screwed up on my initial explanation

There is a clear slight advantage on sharpness using .7, because we're not blurring the luma channel too much. Look closely at small details and you'll see what I mean

-kwag

kwag · #32 04-02-2003, 04:57 PM

@Kane,

However, if you really can't tell the difference on a certain movie, hell

, then drop the temporal filters and use 1.0 for luma

You're getting a huge speed advantage that way

-kwag

dazedconfused · #33 04-02-2003, 04:58 PM

Quote:

Originally Posted by Kane

maybe i´m too stupid to understand, but what should be the advantage of the script WITH temporalsmoother and cleaner? it takes longer to encode, but the image quality is the same, as without temporalcleaner and smoother.

More smoothing = less filesize...the more smoothing you can do without jeopardizing your resulting quality too much, the better. The reason your 2nd sample turned out larger than the first is because you lowered your MergeLuma(Blur) value. Try both scripts again with the same values and you'll see a lower filesize while using temporalsmoother/cleaner.

-d&c

dazedconfused · #34 04-02-2003, 04:59 PM

Beat me by a couple minutes Kwag!!!

-d&c

Kane · #35 04-02-2003, 04:59 PM

ahhh, that was the mistake.
i know what you mean. you can see this effect perfectly, if you look at gimli´s beard in lord of the rings.
the higher the luma value, the less hair you can see, it becomes one brown "thing"

btw:
i did the sample with temporalcleaner-,smoother again with luma 0.7
size: 10,7MB
so, everything´s allright, i´m not crazy *g*

kwag · #36 04-02-2003, 05:03 PM

I guess we can play with mergeluma(blur(.x)), where x can be .5 for shorter films, and increase the value towards 1.0 for longer movies. This way we could probably drop the slow temporal filters completely, and use the merge functions to achieve more compression

-kwag

kwag · #37 04-02-2003, 08:19 PM

Latest test:

Code:

SpaceDust() 
sharpen(0.7) 
mergechroma(blur(1.58)) 
mergeluma(blur(0.7))

Lowered the sharpen just a tad, and I was able to get rid of temporal filters completely.
Has anyone noticed that using this combination of filters, there's hardly any visible artifacts and/or blocks on dark scenes

It looks more like noise, which gives a more natural feel.
Look at this 10 second KVCDx3 sample, encoded with the above script using a CQ of 60

, which is the way this 117 minute full screen movie looks on one CD. It's from the movie "City by the sea": http://www.kvcd.net/city-by-the-sea.mpg

-kwag

Kane · #38 04-02-2003, 08:26 PM

yes, these are mostly the scenes i look at, when comparing samples. those dark areas seem more realistic with this combination. there´s hardly nothing disturbing like bocks in it which destroy the picture

i will test that later, it´s 3:30am, time to go to bed

MrRobot · #39 04-02-2003, 10:08 PM

I used virtually the following script on an avi from a VHS source:

LegalClip
Bilinear .....
mergechroma(blur(1.5

)
mergeluma(blur(0.05))
SpaceDust()
TemporalSmoother(radius=2, strength=3)
TemporalCleaner(ythresh=5, cthresh=10)
STMedianFilter(6,15,4,7,6,15)
DctFilter(1,1,1,1,1,.5,.5,0)
AddBorders....
LegalClip()

The result was very, very blurry. Any good scripts for VHS and 8mm source material?

kwag · #40 04-02-2003, 10:16 PM

Hi MrRobot,

For a VHS cap, try something like this:

Code:

FaeryDust() ( or PixieDust() )
sharpen(0.7) 
mergechroma(blur(1.58)) 
mergeluma(blur(0.7))

See how that works

-kwag