From: Ryan T. <pho...@mi...> - 2004-07-06 03:07:23
|
I was playing one of my DVDs and I noticed some quite choppy output while using the tvtime deinterlacer. When I examined the stream I found that it seemed to be using a somewhat odd pattern of progressive and interlaced flags. The video is primarily interlaced using bottom field first field order but several times it has an odd sequence of three frames. First a progressive frame that has the repeat first field flag set, followed by an interlaced frame that is encoded with the top field first field order, finally there is another progressive frame with the repeat first field flag. After this the video resumes a series of bottom field first frames. I am seeing this with current anonymous CVS version of xine-lib. My setup is as follows: Athlon XP 2200 Radeon 7500 AGP using VIDIX driver Pioneer A07XL DVD 512M RAM Linux with 2.6.5 kernel SBLive using ALSA drivers included with kernel XFree86 Version 4.4.0 The DVD I'm seeing this on is Sailor Moon S Disc 2 during the initial opening sequence(Title 2 Chapter 2). You can find a short (4 meg and around 4 second) sample that shows the effect at: http://www.hamiltoncom.net/~pholan/sailor_moon_op_frag.mpg I understand that this probably can't be fixed until the tvtime deinterlacer has support for tracking the pulldown across progressive sequences but I thought that the sample might be useful to the developers. Also it seems odd that it introduces such noticeable shake. Finally, I have also noticed frequent shifts in field order on some of my other DVDs around progressive sequences. Most recently on Steel Angel Kurumi. I suppose any time there is an odd number of frames using the repeat first field flag the field order has to shift in order for it to display properly on a TV. I guess this is just another example of some of the annoying obstacles MPEG2 encoders can put in the way of good PC playback. BTW, thanks to the all of the xine developers for producing a really good multimedia player. Sorry if anyone see's this twice I originally sent it from the wrong email address. Also in the first copy I forgot to mention I am seeing this with the Greedy2Frame and Greedy deinterlace methods. It is present but less noticable with the Linear method. |