>Is there an existing utility out there that can provide a delay
>between the luma and chroma channels? My camcorder produces video
>with the chroma lagging the luma, with the amount of lag dependent
>(for reasons unknown to me) on the light level. It's really only a
>problem in low light, in which case the lag is quite noticeable
>(several frames). Here there is also the problem of long chroma
>persistence (smearing), but the overall delay is much easier to fix.
That's really strange. Could you say more?
I can try.
I can't really think of a mechanism that would lead to that behavior....
How would the chroma get decoupled from the luma, especially as a
plain delay???
As the ambient light drops, my mostly automatic camera first reduces
the shutter speed and opens the aperture (I think), and then starts to
crank up the CCD gain. But I think it might be doing more than that,
because while the luma just gets noisy, the chroma gets noisy and also
loses detail and shows both what seems to be a net delay of 1-2 frames
and several frames of persistance. You can definitely see the
effects: at the leading edge of motion, the color of the background
winds up on the moving object, while at the trailing edge the color of
the object smears across the background. yuvdenoise makes this even
worse.
My guess is one of two things:
1. Assuming the usual GRGB pixel arrangement, and given that the
chroma depends mainly on the R and B, perhaps there is a physical
mechanism that causes a delay in R and B relative to G at high CCD
gains (seems unlikely, since the only difference between G and RB
is usually a thin filter) The persistance might also be explained
by physics (anyone know about the transient response of CCDs
vs. gain?)
2. The camera may be doing some processing on the chroma in low-light
situations, possibly combining multiple frames in an effort to
improve chroma SNR at the expense of chroma sharpness. This could
explain both the net delay (if they couldn't afford to buffer
several frames to compensate) and certainly the
ghosting/persistence. I guess I could try to call Canon, but I
doubt they'd tell me much.
My analysis of this is complicated by the quite poor chroma quality I
get in low-light situations. It's very washed out, and has poor
resolution even by subsampled standards. I typically apply a gain of
2-3 to the chroma channel to get "normal" color. So it is hard to
decide exactly what is a pure delay and what is a ghost. I did write
a simple y4mdelay as I mentioned (will post later if there is
interest), and found that a delay of 1-2 on the luma helps improve
things. In theory I could equalize the smearing effect with an FIR
filter if it is due to a linear response, but the result would be to
increase the noise even more.
Perhaps I should convert back to RGB, and see if that lends any
further insight (on the assumption that the camera is of the RGB
type).
I made some changes to yuvdenoise to allow independent control of
parameters for each channel, which also helps control these effects.
I passed these on to Stefan, and I think he planned to commit them at
some point.
Dan
-------------------------------------------------------
This SF.net email is sponsored by: eBay
Get office equipment for less on eBay!
http://adfarm.mediaplex.com/ad/ck/711-11697-6916-5
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
