On Wed, 29 Dec 2004, Ray Cole wrote:

Hi!

> I've encoded several things since successfully compiling from CVS last week 
> and it seems to me the quality has noticeably improved.  I originally built 

        Do a cvs update and rebuild now - changes since last week might get
        a (tiny) fraction more quality ;)

> from CVS wanting to try y4mdenoise, but it is just too slow for me to use.  

        It does want a 2+GHz cpu doesn't it?  I had one run on a 2GHz G5 that
        went about 25 hours.  Came out looking beautiful.  I think things will
        look better and run faster when I switch from the ADVC100 to the 
        -300 which can do some denoising on the signal as it's being digitized.

> However I do use y4mspatialfilter right before yuvdenoise, so my chain is 
> {raw video} -> yuvscaler -> yuvcorrect -> y4mspatialfilter -> yuvycsnoise -> 
> yuvdenoise -> mpeg2enc. ..short chain :-)

        Hmmm, the 'yuvycsnoise' filter could probably be dropped - even when I
        was using a Bt878 card that filter didn't seem to make any difference
        (for better or for worse).

        If you're capturing at full frame square pixel (which for a Bt878 is
        640x480) then you are scaling/resampling to 704x480 and not 720x480
        I hope. If the latter then the aspect ratio is being distorted about 2%.
        Full frame broadcast TV (and VHS tapes, etc) in the US are 704x480 with
        10:11 pixels.  This, as arithmetic will show is a 4/3 image equivalent
        to 640x480 with 1:1 pixels.

> Anyway, it appears to me the quality is a little better than before (ie - last
> official release).  Edges seem to be very crisp now.  I'm not sure if that is 

        Glad to hear that the last year of development hasn't been a complete
        waste of time :-) :-)

> from improvements in yuvdenoise, the addition of y4mspatialfilter, improvement
> to mpeg2enc, or some combination :-)  

        "All of the above" <grin>.  y4mspatialfilter didn't exist in the
        last "official release" as I recall though.

> y4mspatialfilter's defaults caused the image to blur really bad, but I was 
> able to tweak the parameters to it to get some good results.  I get a slightly

        Hmmm - interesting.   For capturing from VHS tapes the defaults are
        fairy moderate (-L 4,0.75,4,0.75 -C 4,0.6,4,0.6).  For laser discs I
        bumped up the 0.75 to 0.8.  The -C (chroma) values are definitely
        over-conservative (0.5 would probably have been better) since VHS
        has only about 1/2 the chroma bandwidth that broadcast TV has.

> lower bitrate using y4mspatialfilter, but keep it before yuvdenoise because 
> if I place it after yuvdenoise it tends to blur a little.  Seems like 
> yuvdenoise runs faster than it used to as well, though I've not done any 

        It does.  Stefan and others put effort into speeding it up - nice to
        hear that the improvements are noticeable.

        Cheers,
        Steven Schultz



-------------------------------------------------------
The SF.Net email is sponsored by: Beat the post-holiday blues
Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek.
It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt
_______________________________________________
Mjpeg-users mailing list
Mjpeg-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mjpeg-users

Reply via email to