Log in

View Full Version : CCE Speed x4 adding filters. Where's the bug?


AloneAgain
24th January 2005, 19:52
I think this post doesn't fit exactly in the thread of CEE speedtest, so I post it aside.I appologise if I'm wrong.

The Story:
I've been recently doing some old VHS (PAL anime) to DVD conversions, using CEE 2.67 trial and a very simple AVS script:

Video=avisource("C:\CINE\Erase22.avi").Trim(614 ,163127)
BilinearResize(Video,720,576,6,4,708,566)

_2DCleanYUY2(Video,0,16,2,2,0,16,16) #really bad picture quality

The video was captured with a Hauppage Tuner Card, PicVideo jpeg (Q=19) and a resolution of 720x576.
I was using mostly Doom guide (Thanks a lot ) for CEE encoding parameters so: One pass VBR , Q:60 BitRmin: 1500 BitRMax: 9000, No Quality settings, The TMPgenc CG/animation matrices, and zigzag+progrssive flag. The Encoding speed obtained in a P4 3GHz hyperthreading computer was 0.56-0.60.

But trying to improve the picture quality I added the XSharpen() filter to the end of the script and to my surprise the speed went up to 1.09. Curious I tested others filters and finally found that with the TrBarry's Unfilter (), the speed rose to 2.20 with a better (to my eyes) quality!!
That's for the story.

The Request:
I thought that adding more filters to the processus the encoding speed should decrease, but maybe it's the filters chain order or maybe i'm doing something dumb because my limited experience. Perhaps somebody wiser can enlight me and explain this behaviour deeply.

The Thanks:
Anyway Big Thanks for the Guides, for the great filters and for the help from all you people :D

RB
27th January 2005, 08:30
That's really strange. Unless the PicVideo codec decodes extremely slow, about 2.20 is the speed I expect from a 3 GHz system without any filters. My only guess is that the filters reduced the bitrate required for Q60 so dramatically that it had a real impact on encoding speed - the less data there is to write, the faster CCE encodes.

AloneAgain
29th January 2005, 13:00
RB Thanks for your answer it must be something so, but I 'm still confused . In the meantime I've done some other tests (again VHS Anime, PAL same conditions) and I've got:

1-No filters
avisource("C:\CINE\JAffar.avi").Trim( 2274 ,96778):
Speed=2.32

2-Only Resize
avisource("C:\CINE\JAffar.avi").Trim( 2274 ,96778)
BilinearResize(Video,720,576,8,6,704,564)
Speed: 1.72

3-With Unfilter(0,0)
avisource("C:\CINE\JAffar.avi").Trim( 2274 ,96778)
BilinearResize(Video,720,576,8,6,704,564)
Unfilter(Video,0,0)
Speed=2.09 !!

-4 With Spatial Denoise Alone
avisource("C:\CINE\JAffar.avi").Trim( 2274 ,96778)
BilinearResize(Video,720,576,8,6,704,564)
_2DCleanYUY2(Video,0,16,2,2,0,16,16)
Speed=0.57 :eek:

-5 The Whole Band
avisource("C:\CINE\JAffar.avi").Trim( 2274 ,96778)
BilinearResize(Video,720,576,8,6,704,564)
_2DCleanYUY2(Video,0,16,2,2,0,16,16)
Unfilter(Video,0,0)
Speed= 2.25 !!!!

And the winner is ...;)

PD: I've droped these scripts into VirtualDub and compress with Xvid and the encoding speed decreased as expected.

If I find some spare time I'll do tests with DVD material. :D