[IGSTK-Developers] Tcon minutes March 29, 2007

[Luis Ibanez]

Just a quick follow up on our discussion regarding the performance
of conversion between color systems.

Here is a profile that we got for another project where we are using the
FFMPEG library for compressing and decompressing video sequences.


In  that particular case we have:

1) Image size 1024 x 768 pixels, greyscale, 1 byte per pixel

2) Computer
2.1)  Dell Precision 450 workstation
2.2) with dual processor Intel Xeon at 2.4GHz
2.3) with 3Gb of RAM
2.4) Linux Debian
2.5) Using the gcc compiler with -O3


With this configuration it takes 30 milliseconds to do
the following:

0) Copy the input image into another buffer
1) Convert the grayscale image into YUV,
2) Compress the image using the Flash video codec
3) Decompress the image using the Flash video codec
4) Convert the decompressed image back from YUV to grayscale.


We have not profiled the conversion in isolation, but the
fact that two color format conversions plus video
compression and decompression fit in 30 milliseconds,
give us some indication on how much the color format
conversion alone could take.


     Luis



---------------------------------------------------
On 3/29/07, Patrick Cheng <cheng at isis.georgetown.edu> wrote:
>
> Hi everybody,
>
> Here is the minutes for discussion on VideoGrabber continued from last
> week:
>
> http://public.kitware.com/IGSTKWIKI/index.php/Minutes_032907
>
> Patrick
> _______________________________________________
> IGSTK-Developers mailing list
> IGSTK-Developers at public.kitware.com
> http://public.kitware.com/cgi-bin/mailman/listinfo/igstk-developers
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/igstk-developers/attachments/20070402/b8febaad/attachment-0002.html>