Hi, Nicolas schrieb:
>OK. I did my tests. I used various versions of the mjpegtools, and I >confirm the green tint comes from yuvdenoise CVS. The other tools are >not responsible. > > hmm, this is really strange, as I can not proof any chromatic shifts when using neutral-gray test-charts as well as colored test-charts, may they be noisy or not. I have downloaded your green-tint-version and, well, I personaly wouldn't call that a "strong" tint... It is nearly not observable for me. If it is the denoiser (which could be so -- perhaps my tests to verify that there are no chromatic shifts are not good enough) it only can be a reincarnation of the old rounding problem... (and that could have happen even after I was very carefuly coding arround it... *argh* -> I hope it's not.) >The green tint appears even with low values for yuvdenoise, such as: >yuvdenoise_cvs_20060409 -v 0 -m2,6,6 -t2,6,6 -M2,6,6 > > This seems to vote for some rounding problem inside the denosier... oh, no... >However, I just observe the green tint. I'm not quite sure it's a bug! > > I hope, it is not a bug. I hope it is your material (which could be the case also, as if it has a lot red specles these will be removed as being outliers before denoising). But well, I will check the calculations it does by hand to see, that it really rounds correctly. I really don't want to use floats to avoid rounding errors. Just as an example, what the *old* denoiser did: If the result of the LP was "n.m" the output allways was "n". It never did proper rounding. For the luminance channel this was OK, but for the chrominance channels it ment, that all values drifted towards zero. And as we all know, this is green... This was a clear bug, but was easy to overlook, as it mainly was due to the method the negative numbers in Y'Cr'Cb' are stored... cu Stefan (who now checks the rounding strategies inside the denoiser... again... and hopes to find... nothing -- will also check if it's your material) >I read in a paper magazine video CCD noise is often made of RED pixels. >Wouldn't yuvdenoise remove those red dots, and therefore increasing the >overall green tint? Moreover, even if the encoded "green" videos are >"greener" than the version I got with the "non-CVS" yuvdenoise, I >recognize the picture in the original footage is a bit too red! That's >probably due to those "red dots". > >Is there a way to reduce the green component in the final video, while >still using the CVS version of yuvdenoise? Because that version is >REALLY efficient in removing the noise. I really like the result >(excepted for the green tint). > >Nicolas. > > >------------------------------------------------------- >This SF.Net email is sponsored by xPML, a groundbreaking scripting language >that extends applications into web and mobile media. Attend the live webcast >and join the prime developer group breaking into this new coding territory! >http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642 >_______________________________________________ >Mjpeg-users mailing list >Mjpeg-users@lists.sourceforge.net >https://lists.sourceforge.net/lists/listinfo/mjpeg-users > > -- Gnomemeeting/Netmeeting: callto:ils.seconix.com/[EMAIL PROTECTED] ICQ: 131490319 ------------------------------------------------------- This SF.Net email is sponsored by xPML, a groundbreaking scripting language that extends applications into web and mobile media. Attend the live webcast and join the prime developer group breaking into this new coding territory! http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642 _______________________________________________ Mjpeg-users mailing list Mjpeg-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mjpeg-users
