Menu

#18 Reuse YUV_RGB in tiffhelper

closed-accepted
nobody
None
5
2002-07-23
2002-05-29
P Orye
No

Hello,
perhaps this can serve. It's faster, but I haven't got a
complete overview of what's planned & done.
So should be checked with...

Aim: Reuse code & faster RGB tiff captures with mmx.

YUV<>RGB functions from TimeShift to CSConversion
(Color Space Conversion)

TiffHelper/RGB calls m_YUVtoRGB ~ CpuFeatureFlag

But TimeShift uses RGB(BGRorder), TiffHelper uses
RGB(RGBorder).
So I added "toBGR" flag to C_YUVtoRGB and P3_~.
TRUE -> as before e.i. YUYV > BGR
FALSE -> YUYV > RGB

For the asm functions it is done by selecting different
multiplication arrays: U&V or R&B terms swapped.
Rest is unchanged. (That was easiest way.)

On P3 or better it's much faster.

I could add asm functions for TIFF_CLASS_Y-shuffle and
jpeg YUYV-YUVYUV-shuffle, if they'll remain in use.

Greetings,
Piet

attach: cvs patch + CSConversion.cpp .h
(comments etc not cleaned up)

Discussion

  • P Orye

    P Orye - 2002-05-29

    possible cvs patch + CSConversion.cpp .h

     
  • Rob Muller

    Rob Muller - 2002-05-29

    Logged In: YES
    user_id=429252

    Piet,
    It is good to see you are one of the developers now.
    You do not need my help anymore with applying the patches.

    If you want to know what is done and planned you should ask the developer mailing list.

     
  • Laurent Garnier

    Laurent Garnier - 2002-05-29

    Logged In: YES
    user_id=160731

    The problem is that the used formula seem to be different (and less
    precise) from those I now use in Calibration.cpp and TiffHelper.cpp.
    Please see comments in function RGB2YUV and YUV2RGB with the
    correct formula, especially for the conversion RGB => YCbCr where I
    rounded the coefs only at the last step. I have to reverse formula to
    have more precise coefs for the conversion YCbCr => RGB, that is
    without starting with already rounded coefs.

    So, if you could confirm that your formula give the same result (or a
    better result), why not using them. The idea is to compare the different
    results of the two methods when doing RGB => YCbCr and YCbCr =>
    RGB.

    In the case of still capture, time is less important than precision when
    converting in RGB. IMHO priority is right formula.

    Laurent

     
  • Rob Muller

    Rob Muller - 2002-06-14

    Logged In: YES
    user_id=429252

    Piet,
    We have not heard anything from you the last two weeks.
    Are you still around?

     
  • John Adcock

    John Adcock - 2002-07-23
    • status: open --> closed-accepted
     

Log in to post a comment.