KVCD: What More can I do?

[Jimblob]

Hi

I have an Xvid AVI, 672x288, 1475kbps video and I used script #4 with the second Blockbuster line included. I took out the DCTFiler.
At a CQ setting of 69.95, it fits nicely on one CD. It looks pretty good except for a problem with 2 scenes.
The first is a fade-in to a scene looking out a window being pelted with rain on a grey day. The fade-in has a halo effect.
The second is a dark smoky bar scene. The faces are a bit blocky and mottled.
Any suggestions?

Thanks

[Jimblob]

I forgot to mention that I use AVISynth ver 2.55 and TMPGEnc ver 2.521.58.169 and my Xvid is Koepi ver 1 rc4. I don't use FFDShow.

Thanks

[Prodater64]

Quote:

Originally Posted by Jimblob

I forgot to mention that I use AVISynth ver 2.55 and TMPGEnc ver 2.521.58.169 and my Xvid is Koepi ver 1 rc4. I don't use FFDShow.

Thanks

Hi: If you forget anything, please edit your previous post, if no answer was posted yet.
Maybe if you don't take out DCTFilter, blockiness it will dissapear.
Maybe it is a source matter (bad quality).

[incredible]

Quote:

Originally Posted by Prodater64

Maybe if you don't take out DCTFilter, blockiness it will dissapear.

Depends on the destination resoultion he will encode to!
At 352x288(240) the frequncy cutting effect of DCTfilter will be very recognisable ... as it will soften the encoding very much, ok, by this you will gain compressibility or on the other hand gain quality, means perhaps less blocks will come out, ..... but how will it look that unsharp!? , not mention that unsharp where already 352x288 means everything but not sharp

[Jimblob]

Destination target is PAL 480x576 mpeg2. So far, I have the best results with the following:


AVIsource("D:\AVI\WonderBoys.avi", false)
BlindPP(cpu=4)
Blockbuster(method="noise",detail_min=1,detail_max =3,variance=0.1,seed=1)
Convolution3D(1, 6, 12, 6, 8, 2.8, 0)
GripCrop(480, 576, overscan=0, source_anamorphic=false)
GripSize(resizer="LanczosResize")
Undot()
asharp(1, 4)
UnFilter(50, 50)
STMedianFilter(3, 3, 1, 1)
MergeChroma(blur(1.5))
MergeLuma(blur(0.1))
Blockbuster(method="noise",detail_min=1,detail_max =10,variance=0.3,seed=5623)
GripBorders()


I removed the DCTFilter because of lines in the output. Can you tell me if the order should be changed? What filters can do a better job than the ones being used?

Thanks

Edit: Wow...CQ down approx 15 from MA script which I tried at first.

[Jimblob]

I've made a few more changes. Unfilter is gone. Undot and asharp are before the first blockbuster line. I've added the motion adaptive portion from the latest script after gripsize. My CQ is up to 66. that is just 4 less that the MA script...

AVIsource("D:\AVI\WonderBoys.avi", false)
BlindPP(cpu=4)
Undot()
asharp(1, 4)
Blockbuster(method="noise",detail_min=1,detail_max =3,variance=0.1,seed=1)
Convolution3D(1, 6, 12, 6, 8, 2.8, 0)
GripCrop(480, 576, overscan=0, source_anamorphic=false)
GripSize(resizer="LanczosResize")
ScriptClip(" nf = YDifferenceToNext()" +chr(13)+ "unfilter( -(fmin(round(nf)*2, 100)), -(fmin(round(nf)*2, 100)) ).TemporalSoften( fmin( round(2/nf), 6), round(1/nf) , round(3/nf) , 1, 1) ")
STMedianFilter(3, 3, 1, 1)
MergeChroma(blur(1.5))
MergeLuma(blur(0.1))
Blockbuster(method="noise",detail_min=1,detail_max =10,variance=0.3,seed=5623)
GripBorders()

function fmin( int f1, int f2) {
return ( f1<f2 ) ? f1 : f2
}

Should I change Lanczos to Bicubic? Any suggestions for a higher CQ?

Thanks

EDIT: Any answer would be helpful...any.

[Jimblob]

Thanks to the flood of suggestions , I've worked my way back to the MA script, with the inclusion of the first blockbuster and the C3D lines.

[Encoder Master]

Why you don't use overscan=2, so you can see more in your TV and the encoder give you more CQ.

[Jimblob]

Thank you for the response. An overscan of 2 will give me a 16 pixel border on both sides of my mpeg. Will a higher CQ make up for the lower resolution?

[Encoder Master]

Your TV Cuts 16 Pixel on the left and right side (Every TV) and you see less movie details with this option the picture will be showen central and so the encoder dosen't work with so many movie details.

Look here (Pan scan..):

http://www.incredible.de.tf/

[Jimblob]

With the example I have been using, I have been encoding 6.7% of my video for nothing. What a waste.

Thanks for the link. I will spend a lot of time on that site.