"LanczosResize not only sharpened the
picture, but also enhanced colors. Colors were richer and image
detail was clearer. "

i'm with you black prince,i post a script here (forum)
with LanczosResize (with fluxsmooth,blockbuster ,legalclip)that gives good image.
i test everything (burn and see on tv) and compare the same chapter with differents scripts....
Lanczos is sharp and clear.
:wink:

editing:
oops...post at same time with Sansgrip.

Interesting... I accidentally made a couple of sample strips with the old GOP (max 36), and in the process noticed that using the new GOP and CQ_VBR mode I used the same quality setting to achieve the same file size, but using the new GOP and CQ mode I was able to increase the quality setting by about 4 points.

Edit: Correction, 6 points. From 63.9 to 69.95...

Anyone else notice this?

@SansGrip,

My tests finally narrowed down to LanczosResize and FluxSmooth as
the best combination for CQ 704x480 encodes (in my opinion). They
looked great when viewed on TV. Colours were rich looking and not
washed out as before and FluxSmooth controled the sharpness while
compressing file size. The only glitch was some noticable blockiness
during the water scenes. I increased the CQ to 100 just to see if it
would disappear. It was improved but still there. Scenes other than
water or fire were perfect. Maybe the tweaking Kwag is doing on the
Q-Matrix will help this problem :D Geez testing does payoff :mrgreen:


-black prince

Moving water is, along with fog, fire and smoke, the most notoriously difficult challenge for an MPEG encoder. A DVD-like bit rate is required to represent them sufficiently, and even then sometimes there are still noticible blocks.

I'm glad it does, or I would have wasted hours :D.

It does :D
Try the same thing with this:

However, your CQ curve will be off, so you'll have to re-calculate. Give it a try :wink:

-kwag

26 templates? hehehe ;)

I'm going to try this as soon as my current test is complete (namely a TV compare of CQ_VBR with Blockbuster against CQ without Blockbuster in all resolutions from 352x240 to 704x480).

This is taking forever 8O.

@SansGrip, :)

testing one chapter, 5 minuts with kvcd-lbr
(my player only works in 352x240),
in dvd2svcd prog:

[AVISYNTH_LanczosResize, Fluxsmooth, Blockbuster, Legalclip]
0=LoadPlugin(!FluxSmooth.dll)
1=LoadPlugin(!Blockbuster.dll)
2=LoadPlugin(!Legalclip.dll)
3=LanczosResize(^TargetWidth,^TargetHeight)
4=FluxSmooth()
5=Blockbuster(method="dither",detail_min=1,detail_ max=10,variance=.7)
6=Legalclip()
7=AddBorders(0,^BorderTop,0,^BorderBottom)
!FluxSmooth.dll=C:\DVD2SVCD\Dlls\FluxSmooth.dll
!Blockbuster.dll=C:\DVD2SVCd\Dlls\Blockbuster.dll
!Legalclip.dll=C:\DVD2SVCD\Dlls\Legalclip.dll

8O fantastic! :wink:
the best quality of all my tests!

ps:
using file prediction factor 0.95 ...change this value?
thanks in advance!

Excellent! Have you done a visual comparison with the same settings and script but using method="noise"? I'd be interested to know how much the new method contributes to the quality you're seeing.

I believe the current factor is 1.0.

Sansgrip,thanks about the factor! ( 1.0)

i worked 30 years repairs(don't know better word) televisions,videos.etc.
see television all they long. every day.

for me is the best script at now with a simple test.
very clean! :wink:

i'm waiting for Kwag,black prince and others oppinion too!

Sounds like my kids :mrgreen:.

That's very good news. I hope this new method turns out to be useful.

Hehehe :mrgreen:, tell me about it. Here are all matrixes I know of, and which I tested and compared to the actual "KVCD" matrix. ( Except "Stargate SG-1" Oppps :oops:, I mean "SansGrip SG-1" ) :lol:

:!: :!: 8O

me too!
8O 8O :!: :lol:

Hey guys, don't pay too much attention to all those matrixes. Some of them have been around the net for a very long time. I did benchmark ALL and every one of them against the "KVCD" matrix, which is the one that created the smallest file size while maintained the best quality. The one labeled "KVCD" is the one currently in all KVCD PLUS templates. The BETA-2 is the one I'm currently testing, which has a frequency notch to lower DCT blocks. It should work on every resolution, because the DCT and low lit domains are not (shouldn't be!) resolution dependant. Still, you'll probably get the best results the higher resolution you go. At 352x240 the blocks are just too big, and will always pop-up somewhere, no matter how hard we try to get rid of them :roll:

-kwag

kwag - How long ago was it that you tested TMPGEnc Plus's 2-pass VBR against CQ(_VBR)?

Maybe 6 months ago :!: I even tested CCE, which of course is no good for MPEG-1, so I discarded it immediately. But last time I tested 2-pass with TMPEG against it's own CQ, CQ was much higher quality for the same target size.

-kwag

Are you certain it was the TMPGEnc Plus "new-style" 2-pass, not the 2-pass included with the free version?

Yes, I tested against my TMPEG PLUS. I did use 2-pass NEW. Not OLD 2-pass. I even recall the movie I tested it. It was on Mel Gibson's movie "PayBack".

-kwag

Ok. I was just checking, because one of the first tests I ever did was TMPGEnc new vs. old 2-pass and the difference when seen in BitRate Viewer was like night and day...

Might be worth revisiting, though, with the new GOP and Q matrix etc...

Absolutely :D
I'm currently encoding K-19 at 528x480 with your new "dither" option. So tomorrow, I'll take the final target size and run TMPEG with that size as target for 2-pass VBR. This way we can compare if there's any difference.

Edit: However, there must be a very noticeable difference in quality if we're going to use 2-pass VBR, because the encoding time will be 2X the time compared to CQ or CQ_VBR. So even if it's "slightly" better quality, I think the time to encode wouldn't be worth it ( Specially if the machine crashes 75% in the process :roll: ) :!:

-kwag