[MPlayer-users] Re: mencoder/crop => image distortion

[Jonathan Rogers]

Hans Meine wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> On Thursday 31 July 2003 12:27, Jonathan Rogers wrote:
> 
>>I was trying to end the discussion, but sure enough you showed
>>your true colors. There's nothing worse than a know-it-all (well
>>I don't exactly how much you know, but it's not everything)
>>person.
> 
> 
> Please, before anyone of you posts further mails, could you please calm down? 
> I found the mails of both of you very instructive in the past.
> 
> @Rich: "know-it-all-user" was not nice - and Jonathan is answering many 
> questions on -users, so we all should be thankful. (Even if you might not 
> have answered the same in every case, but people are happy in general if the 
> get good, maybe non-optimal answers, too.)
> 
> @Jonathan: You reacted very aggressively. It's understandable, but words like 
> "asshole" do not really prevent escalation.

Yes, I'm sorry for getting so angry. This was the first time I 
resorted to a personal attack and I'll try not to do it again, 
though Rich does it all the time with far less merit.

> I hope we get nice constructive posts of you two again, soon.
Yes, that is to be hoped for.
> 
> And to the matter of cropping yv12: I would think the facts are clear; yv12 
> has chroma (color) information for two adjacent pixels horizontically. So if 
> an even number of pixels is cropped, everything is fine. Now if an odd # of 
> pixels is cropped, the filters have to create new chrome entries for odd 
> offsets. Those could be the old ones, shifted by one pixel to the left/right 
> (which would make color shift by one pixel "behind" the graylevel image, be 
> technically stupid, and would probably look wrong in _some_ situations, but 
> IMHO would not be apparent on first sight in non-artificial movies) or 
> interpolated entries of both old neighbor entries (meaning a strong smoothing 
> in X-direction, IMHO also not apparent on first sight in non-artificial 
> movies). So if you create a movie with 2-pixel wide vertical color stripes, 
> that would obviously get broken with cropping 1 pixel, but with "normal" 
> movies, one would maybe notice no difference. However, Rich's right that 
> technically, cropping 2 pixels will be much more sane.

Yes I did understood Rich's argument and I never questioned his 
technical understanding or explanation. I was just complaining 
that he wouldn't even try what I described before attacking me 
personally. My argument was not technical, but perceptual. As you 
say, the error may not be detectable, so we may both be right, 
though I'm entirely willing to be proven wrong. As I said before, 
it really isn't worth discussing further and I'll just crop to 
even dimensions since it's not a hardship anyway.

Jonathan Rogers