Hi Andrew > Following up some cool feedback from a kind person who ran some 1.6.2 > mplex/mpeg2enc sequences through a commercial SVCD stream validator I finally > got a lead on the issue with motion artefacts when dual-prime motion > estimation is active. > > It turns out that under certain circumstances with rapid motion present in the > source image the motion vectors generated can be larger than permitted by the > 'f-code' in the picture header. > : > I am about 1/2 done a fix in the current developer branch of mpeg2enc: I've > replaced the old reference encoder derived dual-prime code with a clean > from-scratch routine but it now needs to be tested to knock out the coding > bugs. In the meantime I'll probably put a quick-and-dirty work-around into > the stable branch.
This is an interesting development. Thanks for the heads-up. I guess therefore at this stage you'll hold off grabbing the samples I have until after I've been able to test the revised code? If you still do want the test files, please let me know what filesize I should stick to and I can get them up for you to download. At present the files are between 3 and 7 MB but I could cut that down if necessary. Please advise. Please let me know when a revision is available for testing and I'll run it over my test sequence and let you know what I find. It will be interesting to see whether this addresses the effects I am seeing. One lingering doubt I have is that while the effect was most obvious with rapid movement, there were also clearly artifacts associated with slower motion too. Whether this was in frames which had fast motion elsewhere and whether that fast motion could mess with other areas of the frame I would have to look into in more detail. Of course all this depends on how fast "rapid motion" is. In my case I had a hips-to-head shot of a person who was waving their arms around at a moderate pace and the artifacts were clearly visible associated with the arms. However, at times the camera panned slightly to follow the person and the artifacts were evident along a sharp edge in the background which probably moved no more than 5% of the frame width in 1 second. Regards jonathan -- * Jonathan Woithe [EMAIL PROTECTED] * * http://www.physics.adelaide.edu.au/~jwoithe * ***-----------------------------------------------------------------------*** ** "Time is an illusion; lunchtime doubly so" ** * "...you wouldn't recognize a subtle plan if it painted itself purple and * * danced naked on a harpsichord singing 'subtle plans are here again'" * ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click _______________________________________________ Mjpeg-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/mjpeg-users
