[x264-devel] Re: x264 improvement

[Dmitriy Vatolin]

Dear Champ Yen,

Friday, March 10, 2006, 6:53:00 AM, you wrote:

cy>     I have read the result. It's a amazing and great work!
cy>     I'm glad to see that.

        Not so great, this is initial codec state patch.

        But we think that not so good results of x264 in comparison with
        Ateme on standard sequences like flower, tennis and so on was
        due to this point. So now it's results can be better.

        This result was reached at the end of 2005 (sorry, for delay
        with publication), and our research was continued. In this
        research we collided with problems with accurate codec status
        (state) saving (for future returning to this state). Such
        feature will allows essential increasing of research efficiency.

        I suppose that problem is in speed of current version /
        convenience for research alternative.

        So now we need for cooperation with some other
        researchers/developers to find better compromise ways to solve
        this problem.

        Of course it's possible to make such research on reverence
        version, but sometimes results on JM is good only for papers and
        shows not so good results on practical code so we try to work
        with x264 code to achieve more practical results.

        Ok!

        Any suggestions with solving problem of codec state saving is
        welcome!

        Thank you in advance!

--
 Best regards,
  Dmitriy Vatolin

cy> sorry for my poor English.

    My English is not native also. We can understand each other and this
    is wonderful. :)

cy> On 3/10/06, Dmitriy Vatolin <dmitriy at graphics.cs.msu.ru> wrote:
cy> Dear all, 

cy>   I'm Dr. Vatolin, Graphics & Media laboratory of Moscow State
cy>   University (MSU G&M Lab VideoGroup). VideoGroup in our laboratory
cy>   working in area of video compression also (see for example 
cy>   http://compression.ru/video/ls-codec/index_en.html ). Last autumn we
cy>   have started a small research project "x264 improvement" to analyze 
cy>   x264 sources. Now we have first results!

cy>   We make a big amount of measurements to find where codec can be
cy>   improved (first of all RC of codec). Most code is very good, but
cy>   finally we found such place. We replaced "start_qp" predefined 
cy>   parameter with function of target bitrate. It allows codec to predict
cy>   first frame size more precisely. The improvement is significant at low
cy>   bitrates with relatively short sequences. In our tests we outperform 
cy>   original version of codec (sources 02.12.06) up to 2 Db.

cy>   More detailed information (including modified sources, binaries and examples)
cy>   can be found here:
cy>   http://www.compression.ru/video/x264/x264_improvement_en.html 

cy>   Please use as contacts e-mail x264 at graphics.cs.msu.ru !

cy>   Enjoy! :)

cy> --
cy> Best regards, 
cy>  Dmitriy Vatolin
cy>   Ph.D, Head of Video Group
cy>    mailto:dmitriy at graphics.cs.msu.ru























-- 
Best regards,
 Dmitriy                            mailto:dmitriy at graphics.cs.msu.ru

-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html