On Sat, 7 May 2005, Steven M. Schultz wrote:

(like inside my home) there is a *lot* of chroma noise in the recordings:
On a 100% grey background you can see small patches of slightly red
colored pixels that appear at random and never last longer than 1 frame.

CCD noise :( Small CCDs (consumer camcorders often use 1/6" CCDs, the more expensive ones 1/4.7" and the pro-sumer models 1/3") are quite susceptible to that.

Indeed, it's a consumer mini-DV camera.

So, I tried using 'yuvdenoise -t 1' but it
acted like a digital sledgehammer: The picture looks 100% noise free and
blurry. Even 'yuvdenoise -t 0' is far, far too agressive.

Hmmm, current cvs yuvdenoise does not have a '-t' option that I can see. The usage summary (or looking at the getopt() call in the source module) show:

  INFO: [yuvdenoise]  -Y [n]      spatial Y-filter-threshold
  INFO: [yuvdenoise]  -U [n]      spatial U-filter-threshold
  INFO: [yuvdenoise]  -V [n]      spatial V-filter-threshold
  INFO: [yuvdenoise]  -y [n]      temporal Y-filter-threshold
  INFO: [yuvdenoise]  -u [n]      temporal U-filter-threshold
  INFO: [yuvdenoise]  -v [n]      temporal V-filter-threshold

I did see this new syntax after entering 'yuvdenoise -h' but it did not seem to work either. The reason turns out to be that I entered:


... | yuvdenoise -Y 0 -U 0 -V 0 | ...

to try it and it crashed on that, so I got the impression that this wasn't correct either. Maybe yuvdenoise could check it's input parameters and print a clear error message when the wrong syntax is used, or when the user enters floating point threshold values in stead of integer. yuvdenoise crashes on floating point values as well.. :)

So, I decided to try the new y4mdenoise. This utility did not seem to have
a subtle bit noise remover either. I tried 'y4mdenoise -t 2 -T 6 -z 1 -Z
5' which seems to come close to what 'yuvdenoise -f' could do. These
settings are quite agressive, but less agressive settings fail to remove
the chroma noise.

Actually I think those settings are quite mild and not aggressive at all. "-t 2 -z 1" is usually dones as a "finishing" step since it is so non-aggressive.

Ah. Well, the values I had to use for chroma denoising (6 and 5) are even higher than the manpage recommended values for VHS video, so I was thinking that these were very agressive settings. The subtle bit noise killer 'yuvdenoise -f' of MJPEGTools 0.6.2 was quite effective for removing this kind of noise, so it did not make sense to me why I should need such aggressive filtering to get the same effect with y4mdenoise.


What is the best way to attack this kind of noise in current CVS, and is
there any good replacement for 'yuvdenoise -f' ?

You might try 'yuvmedianfilter' on the chroma only. To disable luma processing use "-t 0" and then experiment with "-T" values to set the chroma thresholding. There is also a "-f" (fast/averaging) option for yuvmedianfilter that may be useful.

I tried that and it is quite effective indeed.

        If the noise is primarily in the darker areas then 'desaturating the
        lows' (removing some of the color info from darker areas) might help
        but I only know how to do that with FinalCutPro.

This sounds like something that is not very difficult to create for MJPEGTools. Is there some sort of 'Hello World' example source out there to learn how to program/compile your own MJPEGTools tool? I do have programming experience, but I don't know a lot about digital video. I might want to play a bit with example source and possibly learn how to implement such a 'desaturating the lows' tool.


Cheers,

Dik


------------------------------------------------------- This SF.Net email is sponsored by: NEC IT Guy Games. Get your fingers limbered up and give it your best shot. 4 great events, 4 opportunities to win big! Highest score wins.NEC IT Guy Games. Play to win an NEC 61 plasma display. Visit http://www.necitguy.com/?r=20 _______________________________________________ Mjpeg-users mailing list Mjpeg-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mjpeg-users

Reply via email to