Dear Anand, Discarding
the chroma samples becomes an application feature i.e. depends on the
application builder how he wishes to perform the same, which will depend on the
computational complexity and memory requirements to perform the same. If we are
capturing YUV422 Interleaved data, then it becomes a bit of a tedious task and
hence I felt that conversion to YUV420 Planar is better. However, if we capture
in Y/C capture mode without interleave, then your suggestion works
fine.
Cheers,
Ganesh
----- Original Message -----
From: Anand K
To: c...@yahoogroups.com
Sent: Thursday, December 23, 2004 12:24 AM
Subject: [c6x] Re: DM642:::video encoder
You are right Ganesh, but usually for the vertical direction, a
subsampling (without filtering) is good enough for most practical uses. In which case, simply discard the color info from alternate fields. The data will still be Edma-ed by the port into the buffers, but
discarded by the user. (very easy to do with the Y/C capture mode, where Y
and C are NOT interleaved)
The chrominance-resampling feature, along
with the 1/2-scaling feature, helps us do the above even with
CIF-resolution. (4-CIF is straighforward: where you discard color info for
every alternate field).
Of course this is not the perfect
approach, but a practical one, in applications with high frame rates where
very high color-fidelity between frames isnt needed.
~ka
--- In c...@yahoogroups.com, "Ganesh Vijayan" <ganesh@e...>
wrote: > Dear Anand, > I guess you are speaking about Sec 3.5.2
of the document. My initial statement that we can't get YUV420
conversion is wrong to some extent. As you have rightly pointed out that
we can use the Chrominance Resampling operation. But in the section, it is
mentioned that sub-sampling happens only in horizontal direction. To quote
the text, "The vertical portion of the conversion must be performed in the software". Hence, we can get only the horizontally sub-sampled version and we need to handle in vertical direction i.e. skipping chroma
samples for alternate lines. So I was partially wrong and partially
correct :O) .. Please correct me if my understanding is wrong. >
Cheers, > Ganesh > ----- Original Message ----- > From: Anand K > To: c...@yahoogroups.com
> Sent: Monday, December 20, 2004 8:53 PM > Subject: [c6x] Re: DM642:::video encoder>
Actually, there is a "chrominance-resampling" feature that can be > used to capture 420 frames from the video ports. >
> Refer document: SPRU629C.pdf >
"TMS320C64x DSP Video Port/VCXO Interpolated Control (VIC) Port > Reference Guide" > > Check out
Section 3.5 on "Video Input Filtering" > > ~ka > > --- In c...@yahoogroups.com, "Ganesh Vijayan"
<ganesh@e...> wrote: > > Dear Suren, > > The output from the Video port will be a YUV 422 frame.
You need to > perform a color conversion which I
think is supported in CSL or can > be written
independently. > > Hope this Helps. >
> Cheers, > > Ganesh >
> ----- Original Message ----- >
> From: suren porwar > > To:
c...@yahoogroups.com > > Sent: Saturday,
December 18, 2004 9:12 PM > > Subject: [c6x]
DM642:::video encoder > > > > > > Hi All, >
> > > I just wanted to know whether the
video encoder(latest or older > version) on EVM DM642
accepts frames in YUV 4:2:0 or do we have to >
resample it to YUV 4:2:0 from YUV 4:2:2. > >
I referred few documents in TI site all the examples, using evm > dm642 accept YUV4:2:2 and then convert it into YUV4:2:0. > > I wanted to know whether Directly YUV4:2:0
could be accepted??? > > If yes
how??? > > > >
Thanks, > > Regards >
> Suren. > > > >
> > > > > > >
|