Re: [libdc1394-devel] Re: single grab DMA example code

[Damien D. <da...@do...>]

Hi all,

Sorry, I must have missed the email on the list...

On Tue, 2005-03-15 at 09:50 +1300, Johann Schoonees wrote:
> Don Murray wrote:
> > So... I developed a couple of simple examples to distribute to PGR 
> > customers back in January.  These examples are for grabbing using the 
> > video1394 device and consequently using DMA transfer.  I am attaching 
> > the 2 examples here.  They are dead simple and are 95% direct 
> > conversions from "grab_gray_image.c".    But I have a few issues I 
> > wanted to bring up.
> > 
> > (1) since this is LGPL, we thought the best thing to do is not to 
> > distribute it from PGR but to simply push it back into dc1394 @ 
> > sourceforge... I'm not sure how to do that or whether you guys even want 
> > this stuff.
> > 
> > (2) I needed to use the "conversions.h,c" file from coriander to do 
> > Bayer tile pattern conversion.  This is some functionality that I think 
> > is quite important for people who want to "acquire images" from a bayer 
> > tile pattern camera.  As such I feel it has a place in dc1394. On the 
> > other hand, it does not have anything to do with IIDC so in that case it 
> > does not have a place in dc1394.  But it doesn't make sense to me that 
> > people who are developing their own image grabbing apps should have to 
> > checkout coriander and yank this code from there and stick it into 
> > their  own dev space... any opinions on where this kind of functionality 
> > should live?  (A new project? Yeah right! :))  IMO it should be moved 
> > from coriander to dc1394... but thats more from a user perspective and 
> > less from a maintainer perspective.
> > 
> > (3) since initially I was planning to be distributing this "from PGR" to 
> > "PGR customers" I mangled all the code to make it look more like our 
> > Windows examples... (hey, I didn't really have a choice ;)).  So a lot 
> > of the variable names have been changed from the grab_gray_image 
> > example.  It makes me wonder if you want to have these kinds of examples 
> > in dc1394 maybe I should take an hour and re-do it from the 
> > "grab_gray_image" basis again and make it match the "look and feel" of 
> > the other examples?
> 
> Don
> 
> I waited a few days for the developers to respond but I did not see 
> anything on the list.  Your example code could be useful and I think 
> your offer was generous.
> 
> (1) How about putting your code in a public part of your company's 
> website and letting me put a link to it in the libdc1394 FAQ?  If you 
> can't do that for some reason I may be able to host it here.

If you're ready to change the code to fit the definitions in libdc (e.g.
bayer patterns) it could be integrated. The style issue is not a problem
for me: it might even help people who are also coming from the MS world.
So my opinion is that as far as it compiles and work without redundant
definitions,... then we can add it in the examples directory.

Other wise we can provide links to your code from coriander, libdc,
camwire, the FAQ,... ;-)

> This may be worth doing if only to lift the Bayer conversion code out 
> of Coriander (with appropriate acknowledgement of Damien Douxchamps 
> and prior authors of course) so that it has higher visibility.

I would be glad to move the bayer code out of coriander. And, in
addition, to move all conversion code out of it too. If necessary we
could start a new lib for color space conversion. Alternatively, it
could find its place in libdc, but I think it would be OT in there too
even though it's a better place. 

> (2) Camwire may be a good home for the Bayer conversion code, being 
> just one layer up from libdc1394.  Patches are always welcome.

IMHO either we think that the bayer code is very generic in which case a
new lib is the way to go, or we think it's somehow related to IIDC and
then libdc is the choice.

It would be a great value for the image processing community if we can
find the manpower to build a good optimized lib for color space
conversion (by optimized I mean MMX/SSE/SSE2...). Maybe we don't need to
start from scratch and the bayer code could be integrated in an existing
project...

> (3) Your code looks perfectly servicable to me.  I would rather see it 
> openly available as it is than held back over concerns of style.

Definitely.

Damien

-- 
  _    Damien 'Takahara' Douxchamps
 ('-   Post-doctoral investigator
 //\   Image Processing Group, Nara Institute of Science and Technology
 V_/_  http://chihara.aist-nara.ac.jp/