You're welcome. The caveats would be that this is obviously just the _average_ milliseconds required to decode a frame -- there still could be significant variation between I, P, and B frames, or other variations within the stream.
And getting more data points (with a wider variety of bitrates) would probably be a good idea to better characterize the actual correlation. I don't mean to suggest 87% is a serious figure given that it came from five random MPEG-2 files on my hard drive. I think any claim that it scales linearly with luminance sample count (which is another way of saying linearly with macroblock count) would also have to clarify -- would the number of bits per macroblock stay constant as you increased the number of macroblocks, or would you keep the total number of bits per frame constant, or what? But hopefully these ballpark figures will be helpful to you. Best, Keith On Fri, 29 Feb 2008, Krishna Prasad wrote: > Hi Keith > > Thanks for your earliest reply. > I will look in to the details of whatever you had sent.If I have any doubts, > I will keep you posted. > > Regards > Krishna Prasad > > > On Fri, Feb 29, 2008 at 5:24 PM, Keith Winstein <[EMAIL PROTECTED]> wrote: > > > Hi Krishna, > > > > I did some experiments on a Pentium 4 at 3 GHz with libmpeg2 0.4.1, using > > some MPEG-2 video streams I had lying around of various resolutions. Here > > are the results: > > > > Luminance resolution -- speed > > > > 720x480 -- 319 fps > > 1280x720 -- 124 fps > > 1280x720 -- 135 fps > > 1920x1088 -- 98 fps > > 1920x1088 -- 86 fps > > > > If we take "frame size" to mean "total luminance samples," then we get > > these data: > > > > Luminance resolution -- milliseconds per frame > > > > 345600 3.1 > > 921600 8.1 > > 921600 7.4 > > 2088960 10.2 > > 2088960 11.6 > > > > For these data, I get an R^2 value (a test of linearity) of 0.87. Where 1 > > is a perfect correlation (linear relationship) and 0 is no correlation. > > > > So, I would say it's basically 87% linear. Not perfect, not always true, > > but a good approximation. > > > > Best, > > Keith > > > > On Fri, 29 Feb 2008, Krishna Prasad wrote: > > > > > Hello Keith > > > > > > Thanks for your reply. I will have a look through it. > > > > > > Many of the paper references state that the decoding time varies > > linearly > > > with the Frame Size. Does the policy holds good always? > > > > > > Can I get your feedback on this particular question?. > > > > > > Regards > > > Krishna Prasad > > > > > > ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Libmpeg2-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmpeg2-devel
