KVCD: Trouble with gibbs effect?
 

Hi all. I'm making my first KVCD from a DVD rip (as opposed to VHS capture), and I'm having a really hard time with Gibbs effect on this one. The source is "American Ninja" DVD, which is 95 minutes 16x9, and as you can guess from the title it's a pretty active movie. The DVD source is a little on the dirty side - not the best transfer I've ever seen, but it's certainly within the bounds of acceptability.

I thought surely the 352x240 template would be the way to go for such an active movie, but even this template has a hard time fitting it on one CD at decent quality. I used file prediction to come up with a CQ_VBR of 34, but at that quality I see Gibbs effect EVERYWHERE in the picture, even on still scenes. 8O It's awful.

I have made KVCDs in the past from worse sources at lower CQ and not seen Gibbs effect like this. I have tried different combinations of AVISynth filters, but no luck so far. My first script was like this:

mpeg2source("Z:\AMERICAN_NINJA\american ninja.d2v")
Crop(4,10,-4,-10)
BicubicResize(336,336,0,0.6,0,0,712,462)
FluxSmooth()
BB_Resolution = 352*240
BB_StrengthConstant = 352 * 240 * 20 # Base strength
StrengthValue = round (BB_StrengthConstant / BB_Resolution)
Blockbuster( method="noise", detail_min=1, detail_max=10, variance=1 )
Blockbuster( method="sharpen", detail_min=20, detail_max=90, strength=StrengthValue)
LegalClip()

When I saw the result, I thought maybe Blockbuster was the culprit, so I decided to try stronger temporal smoothing to improve compressability. I can live with a few blocks if I can get rid of the gibbs.

mpeg2source("Z:\AMERICAN_NINJA\american ninja.d2v")
Crop(4, 10, -4, -10)
BicubicResize(336,336,0,0.6,0,0,712,462)
Convolution3D (0, 6, 10, 6, 8, 5, 0)
TemporalSmoother(2)
LegalClip()

Same result. Both of these scripts produce nearly visually perfect AVI files, but the MPEGs just look like pure crap. I would expect to see some gibbs on the very high action scenes at a quality of 34, but not this much and not on the still scenes. :roll:

Any suggestions?