Hi -

> From: Florin Andrei <[EMAIL PROTECTED]>
> >     Did you denoise the material first?   The average bit rate shown
> 
> No.

        Try it - as an experiment if nothing else.   

> My source is a rather cheap DV camera (Sony TRV240). On one hand, being

        I can go you one cheaper - I've a Sony TRV120 [it's an old Digital8
        unit probably no longer made] that was $99 as a refurbished
        unit :)   Works a lot better than I expected.  And so far, the Xmas
        family video footage is coming out looking very very good considering
        it was done with ambient lighting only.

> being cheap has its own disadvantages: the CCD has a rather nasty color
> noise when filming in darker places.

        CCDs have gotten better since the 240 (and 120) were introduced.  The
        740 or 840 is state of the art as far as single CCD miniDV units go.

        Beyond that there are the 3 chip CCD cameras that have better low-light
        capability but they're fairly expensive.

> I'm not sure what to do. I don't want to apply denoise filters blindly.

        Experimentation is the key.   You're already tossing away a _lot_ of 
        the information by going to SVCD - might as well make the best use
        of what few bits the format has to offer.

> I know precisely how the noise looks like, but i don't know which filter
> works best against it (if any).

        'yuvdenoise -S 0 -l 1' is a very NON aggressive (fast reacting) - almost
        mild one could say - filter.    You can make it even less aggressive
        by specifying a higher threshold (default is 5) with "-t 8".

> >     original mail item with the mpeg2enc command) use -N to mpeg2enc?
> 
> No.

        Try it on one capture and see what happens.  It, as the mpeg2enc's
        help says:

--reduce-hf|-N
    Reduce high frequency resolution - useful as a mild noise reduction

        To paraphase the old joke (which was about programs growing large):

        Q: How do videos get large
        A: One kbit/sec at a time :)

        If you can save 4 or 5% and not notice the difference (especially
        on a TV) then I'd call that Success.   You _might_ even achieve that
        order of savings just by using y4mscaler instead of transcode's scaling

        Is it possible that transcode's scaling is introducing high frequency
        noise that's causing the encoder to use more bits?   If so then 
        switching scaling utilities only and not using other filters might
        be a good experiment to try.

> So, this is like a low-pass filter? (i'm much more familiar with the
> audio stuff)

        That's another way to look at it :)

>Wouldn't it cut off the fine details?

        You're already doing that - or put another way, with only 2500Kbits/sec
        for 480x480 there aren't enough bits available to hold all of that
        detail.   That's what DVDs and the ~9-10Mb/s bitrates are for <grin>.

        Good Luck!

        Steven Schultz
        [EMAIL PROTECTED]


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users

Reply via email to