Log in

View Full Version : QS deinterlace - pIII/pIV


Wilbert
3rd April 2002, 09:35
A small request for pIII/pIV users:

Recently someone pointed to an optimized version (for pIII/pIV only) of smart deinterlace. The corresponding avs-script is given below.

I would be glad if some pIII or pIV user could test this script since VDub crashed on my Athlon using that script and I want to add it to the collection of scripts of vdub-filters. (pIV users have to change "\QSdeinterlacePIII.vdf" in "\QSdeinterlacePIV.vdf")

I would like to know the following:

1) Does the script work and does it give the same results when using the corresponding vdub-filter?

2) Does it give the same results as the smart deinterlace filter and how much is it faster?

script + place where to download the QS deinterlace filters:

####################################################################
# QS deinterlace by ?, v? (optimized version of Smart_Deinterlace, #
# for the pentium III users) #
# #
# http://freetime.sinor.ru/observatory/article.phtml?qsdeinterlace #
# Please use http://babelfish.altavista.com for translation. #
####################################################################

function VD_QSdeinterlace(clip clip, string "mode", string "channel", bool "show_motion", bool "blend",
\ bool "cubic", bool "motion_map_denoising", int "m_threshold", int "sc_threshold",
\ bool "fs_before", bool "shift", bool "fs_after", bool "disable")
{
LoadVirtualdubPlugin(VirtualDub_plugin_directory+"\QSdeinterlacePIII.vdf", "_VD_QSdeinterlace", 1)
mode = default(mode, "frame")
mode = (mode=="frame") ? 0 : (mode=="field") ? 1 : (mode=="both") ? 2 : -1
Assert(mode>=0, """VD_QSdeinterlace: "mode" parameter must be "frame", "field", or "both"""")
channel = default(channel, "luma")
channel = (channel=="luma") ? 0 : (channel=="all") ? 1 : -1
Assert(channel>=0, """VD_QSdeinterlace: "channel" parameter must be "luma" or "all"""")
return clip._VD_QSdeinterlace(default(show_motion,false)?1:0, default(blend,false)?1:0,
\ default(m_threshold,15), default(sc_threshold,100), default(shift,false)?1:0,
\ default(fs_before,false)?1:0, default(fs_after,false)?1:0, default(disable,false)?1:0,
\ default(motion_map_denoising,false)?1:0, mode, channel, default(cubic,true)?1:0)
}


# example1 VD_QSdeinterlace("frame","luma",false,true,false,false,15,100)

# example2 VD_QSdeinterlace("frame","luma",false,false,true,false,15,100)

# example3 VD_QSdeinterlace("field","luma",false,false,true,true,15,100)

Victoire
3rd April 2002, 14:26
I have PIII-1133. So QSDeinterlace (PIII-version) works fine. 0.5b-version produces the same output as SmartDeinterlace, but 0.52b has changes in algorithm so it differs. IMHO 0.52b has more sharp edges than Smart and slightly laerger area for treating.
I've noticed that this plugin crashes sometimes with DivX 5.0 but Smart was also crashes with it. I don't know why. These errors are non-reproductible and occur when system does some work.

Wilbert
3rd April 2002, 14:44
"I have PIII-1133. So QSDeinterlace (PIII-version) works fine. 0.5b-version produces the same output as SmartDeinterlace,"

Nice, thank you very much! Did you also look whether the script of QS-deinterlace is really faster than the script of smart deinterlace, and how much?

"but 0.52b has changes in algorithm so it differs. IMHO 0.52b has more sharp edges than Smart and slightly laerger area for treating."

Ok. (I didn't look at this filter, I guess the script will be the same.)

"I've noticed that this plugin crashes sometimes with DivX 5.0 but Smart was also crashes with it. I don't know why. These errors are non-reproductible and occur when system does some work."

I will add a warning to the script about DivX 5.0.

Any takers for the pII and PIV versions?

Victoire
3rd April 2002, 16:04
"Nice, thank you very much! Did you also look whether the script of QS-deinterlace is really faster than the script of smart deinterlace, and how much?"

Test:
video film (about 4 minutes), no audio
Filter options:
Frame&Field, Blending, Denoising
QSDeinterlace ver. 0.52b 8:12 minutes
Smart 2.7b2 10:22 minutes

So, speed profit is about 26%.

Today no errors occured when I use QSDeinterlace + DivX 5.0 with 2-pass compression. I guess errors were due to GMT option in DivX 5.0

frank
5th April 2002, 13:49
QS deinterlace works in RGB space only!

MPEG conversion in YUY2 color space with

DVD2AVI -> AviSynth (MPEG2DEC.DLL, Smartdeinterlace from Dividee,
Bicubicresize) -> TMPGEnc, CCE

is 50 % faster than QS deinterlace!!
Tested on PIII 800.

And framing with ConvertToRGB -> VirtDub will also be much faster than the RGB mode with QS deinterlace.

Victoire
5th April 2002, 17:27
Well, with AviSynth faster but quality loose...
Plus authors of QSDeinterlace promise that they will make filter better and faster.
I prefer make my home video with QSD - they are really better (IMHO).

---

tested both on PIV - practically no differences by speed but differences in quality of picture: to my mind QSDeinterlace (and Smart, but it's slower) is better in quality. I wait for new beta-versions. And authors want to make their filter for AviSynth!!!!