PDA

View Full Version : Am I capturing correct/best way?


ThNoctus
17th July 2002, 14:57
Let say I'm using my Hauppauge PCI capture/tuner card in combination with my digital satellite box for video capture for almost two years. I connect these two devices with coaxial cable (I also could use s-vhs connection but at this moment that isn't so simple; if I found out that there is better quality possible with latest configuration I'll do that because my main priority is quality).
Usually I capture with avi_io or virtualVCR with picvideo codec or huffyuv in yuv2 format at 576x576 or sometimes 768x576 (it depends on driver: hauppauge vfw 3.11 or btwin 5.3.0 or hauppauge wdm drivers) but sometimes I have serious problems with deinterlace. Also some high motion scenes aren't perfect.
I know there are many topics about deinterlace artifacts so I don't want talk moreabout that instead I need suggestion about basic setting with picvideo or huffyuv codec - which one is better, which format (rgb, yuv) is optimal and my enigma is 2 field setting in picvideo (I found sometimes that is good setting: 2 fields if >288 lines for PAL, but without explanation)
I'm using Vdub with different combination of smart/area/MAP-deinterlacing techniques. I have sometimes very good results with dynamic field order correction filter.
Now I'm gonna do my tests and research with avisynth related deinterlacing.

Also I need opinion about Cyberlink PowerVCR / Intervideo WinDVR MPEG2 capture solutions (I have 1.53 GHz on A7V133, 256MB SDRAM Hyundai).

Best regards
:rolleyes:

Fasola_Jan
17th July 2002, 21:11
Hi :)

About enigma:
PAL video is two fields per frame, so
576/2=288

NTSC video is two fields per frame, so
480/2=240

PicMJPEG codec supports interlacing, so with this setting
You are telling codec to capture two fields (not one
big combined frame). Result - less blending between lines.


More about interlacing:

http://www.100fps.com


S-Video should be better (but not for long distances)


Bye.
Fasola

DaveQB
20th July 2002, 05:38
which is better to capture with

RGB24 or YUY2 for quality ??

i have 80 gig's to capture my 40+ minute long videos so space isnt an issue.

it appears YUY2 has better compression therefore i am thinking i should use RGB 24 for better quality, am i right ??

theres doesnt appear to be anything in it when i play the two different (RGB 24, YUY2) types back :confused:

Fasola_Jan
20th July 2002, 08:21
Hi DaveQB :)

Most of consumer level capture boards, captures internally in YUV,
so by capturing in YUV You are avoiding color space conversion
to RGB (such conversion, if improperly done, may introduce
some minor color distortions).

BUT

In terms of preserving color information, RGB is better
than YUV(4:2:2)

Pls, read section:
"Which capture format should I choose if I'm going to use video compression?"
from VirtualDub docs:

http://www.virtualdub.org/docs_capture


Bye.
Fasola

Dat
20th July 2002, 08:43
Originally posted by Fasola_Jan
Hi :)

About enigma:
PAL video is two fields per frame, so
576/2=288

NTSC video is two fields per frame, so
480/2=240

PicMJPEG codec supports interlacing, so with this setting
You are telling codec to capture two fields (not one
big combined frame). Result - less blending between lines.


More about interlacing:

http://www.100fps.com


S-Video should be better (but not for long distances)


Bye.
Fasola

Are there any loseless codecs that support interlacing? :D

Also, i captured some video of me at a car show, and on some shiny things (ie the cars) the colour looks as though its flickering, anyone know why this is?

Thanks,

Dat

Fasola_Jan
20th July 2002, 09:46
Hi Dat :)

Q: Are there any loseless codecs that support interlacing?

A: I do not know any... BUT is it really needed? When it is loseless
there should not be blending between lines because of compression.

Q2: flickering color
A2: I have no idea. More info pls.
Also there is VirtualDub filter "Antiflicker"
made by famous Donald Graft:

http://sauron.mordor.net/dgraft/flick.html

Maybe You will find it useful.


Bye.
Fasola

Dat
20th July 2002, 19:01
Originally posted by Fasola_Jan
Hi Dat :)

Q: Are there any loseless codecs that support interlacing?

A: I do not know any... BUT is it really needed? When it is loseless
there should not be blending between lines because of compression.

Q2: flickering color
A2: I have no idea. More info pls.
Also there is VirtualDub filter "Antiflicker"
made by famous Donald Graft:

http://sauron.mordor.net/dgraft/flick.html

Maybe You will find it useful.


Bye.
Fasola

Q2:

It's hard to explain, its not really flickering, it kinda looks as though its vibrating or something, do you think it could how i captured the video? but the thing is it looks fine in AVI format.

Thanks for the advice :)

Belgabor
20th July 2002, 19:46
Originally posted by Fasola_Jan

Q: Are there any loseless codecs that support interlacing?

A: I do not know any... BUT is it really needed? When it is loseless
there should not be blending between lines because of compression.



To improve the comperession ratio? I still wonder if HuffYUV could offer better ratio if it was interlaced enabled. (I'm thinking of e.g. placing the two frames besides eachother internally)

Regards
Belgabor

Dat
20th July 2002, 19:56
As regards to my flickering problem, it only happens when i encode interlaced for some reason, so i guess that is my problem right there... hmmmmm, how to fix it???

Dat

Fasola_Jan
20th July 2002, 21:57
To Dat : "pure" colors are very hard to encode in any
high-compressed video format. Try to "legalize" them - reduce
chroma range from 0-255 to 8-235 (TMPGEnc/MPEG Setting/
/Quantize matrix Tab-> disable option "Output YUV data as
basic YCbCr not CCIR601".

To Belgabor : VERY good point. Two small, less complex pictures
should compress better than one big, with more complicated structure.


Bye.
Fasola

Dat
21st July 2002, 17:33
Originally posted by Fasola_Jan
To Dat : "pure" colors are very hard to encode in any
high-compressed video format. Try to "legalize" them - reduce
chroma range from 0-255 to 8-235 (TMPGEnc/MPEG Setting/
/Quantize matrix Tab-> disable option "Output YUV data as
basic YCbCr not CCIR601".

To Belgabor : VERY good point. Two small, less complex pictures
should compress better than one big, with more complicated structure.


Bye.
Fasola

Just tried it, and does look better, but it still doesnt look right, maybe it was the colour settings when i captured the video.

How is it possible to optimize my settings? is there a site or guide or something?

Thanks,

Dat

Fasola_Jan
21st July 2002, 21:20
Very good video capture tunig guide from great
LukasVideo site:

http://www.lukesvideo.com/videotuning.html

Bye
Fasola