On Wed, 26 May 2004, Dan Scholnik wrote:
> Doesn't the ' in Y' indicate that the digital data has been
> gamma-corrected to compensate for the nonlinear CRT response? In that
I'd guess so - what does one get (Y or Y') when running a S-Video
cable from a VCR to a Canopus box and then via IEEE1394 into a
system?
> Out of curiosity, does the noise of the uncompressed recording look
> worse in black frames than regular ones, or is it only after mpeg
> compression that things look bad? If the latter, than perhaps you want
The latter. I see the blocks/splotches all the time in dark scenes.
Also during fades in from black and then fades to black.
It's only after mpeg compression that it looks bad (sometimes very
much so). Using various filters (such as yuvmedianfilter and -D 10 to
the encoder helps a little bit but it's not enough.
> to try to filter with the encoder in mind, possibly testing and
> conditionally filtering the video in 8x8 (or is it 16x16?) blocks. I
That is the plan - I was thinking of offering a choice of 8x8 or
16x16. I can't see a need to offer less than 8x8 or arbitrary
sizes.
For example here's what a few Y entries from a black frame look like
(values are decimal):
0000200 030 026 022 023 030 026 024 026 027 025 025 027 027 030 026 023
0000220 023 025 025 026 032 030 024 024 027 026 023 024 031 023 022 027
0000240 030 021 022 030 030 025 026 027 026 025 026 030 031 024 030 027
0000260 024 025 030 024 025 027 030 027 023 026 027 024 026 031 024 025
0000300 027 025 024 024 032 026 024 024 027 024 024 027 026 023 030 025
0000320 024 026 025 026 027 031 022 022 026 030 024 025 027 026 020 026
0000340 031 025 025 026 027 024 024 027 026 025 027 025 024 023 030 026
0000360 021 025 030 026 022 022 031 031 025 023 025 027 023 023 030 027
To the eye that looks black, but the encoder sees the constantly
changing (by small, usually +/- 2) values. Comes out looking like
alternating blocks of darkgrey/black.
> Temporally stationary CCD noise, AKA dark noise, can be subtracted out
> of long exposures (usually greater than 1 second) to improve low-light
> SNR. At shorter exposures, such as 1/30 or 1/60 of a second,
> time-varying noise usually dominates and dark frame subtraction would
The camera (which was a still camera, not movie/camcorder) claimed
it did it for the shorter exposures - but perhaps it was just
advertizing hype.
> only make things worse. At best you could subtract out the mean if
> "black" has a constant bias. Besides, since dark noise doesn't change
That's what some captures appear to have - the black levels are 4 to
6 units high from the looks of it. Perhaps I'll add an option to
specify an offset to be uniformily subtracted from the luma value.
I'll putter around with things and see what I can come up with.
Cheers,
Steven Schultz
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g.
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
